Closes the COMPLETED-lie where a freestyle stage passes by producing a
schema-valid artifact regardless of whether its build actually succeeded:
a verify stage could run 'npm run build', watch it fail, then self-attest a
build_verified artifact and transition to done. Tool failures never failed
the stage.
The compiler now flags the terminal stage autoBuildGate=true when no stage
declares an explicit build_expectation. Code-ness is decided at runtime, not
compile time (the declared kind is always the generic file_written; only the
written paths reveal code): runExecutionGate folds the FileWritten manifest via
sessionProducedBuildTarget() and promotes to a PROJECT build when a build
manifest (package_json/gradle_module) or an imports_resolve-bearing path was
written, then runs the real command from .correx/project.toml [commands]. A
non-clean exit fails the stage retryably; a docs-only plan is left untouched.
Live-proven (run 3, session 28812b44): gate fired on the real
'npm --prefix frontend run build', exit 2 on a bad tsconfig, RetryAttempted
instead of a false COMPLETED.
Modules green: transitions+workflow 67, kernel 60.
Replaces the single shared per-stage retryCount (reset on TransitionExecuted,
shared across all post-stage gates) with a per-gate budget. Motivated by run
2e468f9f, where a promising freestyle run was terminally FAILED because the
contract gate burned all 3 shared retries on one trivial miss before the
semantic-review gate ever ran.
- StageExecutionResult.Failure gains a `gate` id; each gate tags its failures.
- OrchestrationState gains gateRetryBudgets / gateFailureFingerprints /
gateSalvageUsed (all rebuilt from events, reset per-stage on TransitionExecuted).
- FailureFingerprint normalizes+hashes the failure reason; RetryCoordinator.decide
charges a gate's budget only when the fingerprint is unchanged (no progress),
so a moving run is never penalised. Returns Retry | Exhausted.
- Hybrid exhaustion: deterministic gates fail; the review gate consults a
SalvageJudge (LLM, or a deterministic allow-one-reset fallback), recorded as
RetrySalvageDecidedEvent (invariant #9). CONTINUE resets the gate budget once;
a second exhaustion is terminal.
- Review gate is now the sole authority on review-retry termination; the old
REVIEW_BLOCK_RETRY_CAP is repurposed as a high absolute backstop only.
- ServerModule escaped-exception path now records a truthful terminal
WorkflowFailed (real stage/reason/retryExhausted) via recordUnhandledFailure.
Tests: DefaultRetryCoordinatorTest (fingerprint charging) +
GateRetryBudgetExhaustionTest (deterministic exhaust / review CONTINUE-then-
terminal / review FAIL / null-judge fallback) + reducer coverage.
Plan-compile gate (#19): PlanCompilationCheck seam (null-on-replay),
runPlanCompileGate guards a produced execution_plan slot and compiles its
cached content via ExecutionPlanCompiler, emitting PlanCompileCheckedEvent.
On failure the architect gets a retryable hand-back and self-corrects;
exhaustion terminates the dead-end workflow instead of shipping an
uncompilable plan.
Semantic-review gate (#20): SemanticReviewer seam + StageConfig.semanticReview
opt-in (TOML semantic_review + freestyle plan JSON). runReviewGate reads the
FileWritten manifest (never blind context), asks a review-capable model for
PR-comment findings, and is advisory — blocks retryably only on FAIL plus a
correctness finding >=0.7 confidence, capped at 2 blocks so a stubborn stage
completes advisory-only rather than trapping. SemanticReviewerImpl talks to
the provider directly; any reviewer error degrades to PASS. REVIEW_MAX_TOKENS
is 8192: gemma is a reasoning model and 1024 truncated it mid-reasoning before
it reached the JSON verdict (empty content -> parse degraded to PASS).
Both gates live-QA'd on the real repo: plan-compile passed on a freestyle
web-UI scaffold; semantic review caught all 3 planted bugs in a maxOf() and
exercised the block->retry->cap safety valve end-to-end.
Gate 2 (contract) and Gate 4 (execution) of the 4-gate staged-verification
design, plus the two gate-quality fixes found in live QA.
- Gate 2: KindContractTable (kind→assertions registry, universal floor,
additive project overrides) + KindInference (path→kind, pure/replay-safe)
+ ContractAssertion model; ContractGateEvaluatedEvent (registered);
ContractAssertionEvaluator seam + FileSystemContractEvaluator (FS/TEXT,
COMPILER skipped); runContractGate in runPostStageGates — a stage that
produces file_written artifacts must have every declared+written file
exist/non-empty/parse, else retryable hand-back with per-assertion feedback.
- Option A: ExecutionPlanCompiler puts concrete (non-glob) plan `writes` into
StageConfig.expectedFiles, so a declared-but-unwritten file fails file_exists.
- Gate 4: BuildExpectation enum (none|module|project|tests) + plan field +
closed-set compiler validation; runExecutionGate resolves the alias against
the bound project profile commands and runs it as a deterministic gate.
- Live-QA fixes: react_entry kind so main.tsx/index.tsx aren't required to
export a default component (was an unwinnable false-positive); JSONC
tolerance (allowComments/allowTrailingComma) so tsconfig.json passes valid_json.
Seams are null on the replay harness (no-op) — recorded events are truth (#9).
Context rework from the 2026-07-05 context-poisoning diagnosis (session
6ca236c4 read-loop):
- ContextBucket on ContextEntry: REQUIRED (task, constraints, needed
artifacts, failure feedback) is never pruned/compressed/dropped; the
build fails fast (RequiredContextOverflowException -> non-retryable
stage failure) if required alone exceeds the stage budget.
- OptionalCeilings: per-sourceType token caps for the optional block
(repoMap 1200, relevantFiles 1600, journal 800, profiles 400/600,
vocabulary 400), tuned for the 9-26B tier at 16k; enforced by
truncation (journal keeps its tail, others their head).
- Repo map is now a structural 'what exists' listing (alphabetical,
per-directory, no recency ranking, no symbols); .md/docs paths are
gated behind an explicit docs mention in the stage prompt — retrieval
hits remain the only other path for docs to enter context.
- file_written needs render as a manifest of ALL files the producing
stage wrote (projected from ArtifactContentStored -> ToolInvocation
-> FileWritten events), fixing the last-write-wins collapse; content
stays lazy behind file_read.
- Decision journal folds a stage's RETRY records into one summary line
once a later transition leaves that stage (render-time only).
- Dedupe retrieved-file lines in buildRelevantFilesEntry.
Deterministic repair pipeline (extraction, classification, policy, gated
LLM-repair rung) with ArtifactRepairAttempted/Failed events. Also fix
JsonSchemaValidator: a schema with no declared properties describes 'any
object' — unknown-property enforcement now only applies against a declared
shape (empty-properties schemas were rejecting every real artifact).
RepoMapComputed recorded per session (in-memory scan cache keeps the walk-once
guarantee); zero-score embedder hits filtered before reaching the agent, falling
back to the deterministic repo map; REFERENCE_EXISTS surfaces a closest-path hint
via closestPathByFilename. Breaks the hallucinated-package-path read loop even
with a noop embedder.
Three tool-path correctness fixes:
- Tool-execution events were emitted twice — SandboxedToolExecutor and
SessionOrchestrator.recordToolExecution both emitted Completed/Failed/
FileWritten per call (double TUI rows, double-counted metrics). Split
ownership: orchestrator is the sole recorder of Completed/Failed (truncates
output, drives the read-before-write gate); executor owns Started +
FileWritten (pre/post-image hashes it alone can capture).
- computeToolPreview still gated on operation == "write", but file_write lost
its operation param when delete was split out, so it bailed to raw-args for
every write (approval card showed raw JSON). Drop the dead guard.
- A file_read blocked by REFERENCE_EXISTS can never complete, so it could never
lift read-only mode: a stage writing a NEW file deadlocked (writes filtered,
every read of the not-yet-created target blocked) until MAX_TOOL_ROUNDS.
Lift read-only on a REFERENCE_EXISTS block, and reword the block message to
tell the model to file_write directly. Regression test added.
DefaultSessionReducer skipped WorkflowCompletedEvent to avoid terminalizing on freestyle planning's mid-session completion, but that left execution/normal completions stuck ACTIVE forever — headless watchers and approval loops never saw a terminal state. Add workflowId to WorkflowCompletedEvent (defaulted for replay compat), pass graph.id at all completeWorkflow sites, and terminalize to COMPLETED unless it's the freestyle_planning handoff. Correct stale tests that pinned the ACTIVE-forever behavior.
Uncommitted work from the session-robustness-and-dox branch sweep
(docs/qa/QA-session-robustness-and-dox.md), verified alongside the
compression/context fixes:
- SandboxedToolExecutor: validate tool args centrally before dispatch. A
malformed/missing-arg call becomes a recoverable ERROR: (surfaced with the
tool's arg schema so the model can correct + retry) instead of stranding
the stage with no artifact. + validation test.
- PlanLinter: seed artifacts (analysis) produced by the planning phase count
as available producers, so a plan stage that `needs` them isn't flagged as
an unproduced-need; H1 unproduced-needs + trap-state checks. + tests.
- DefaultSessionOrchestrator: live-QA robustness fixes (event-tail /
per-stage budget + retry handling).
- workflow prompts/configs: DOX AGENTS.md alignment + freestyle/task/role
prompt tweaks.
- SessionOrchestratorIntegrationTest: coverage for the above.
- FreestylePlanningWorkflowTest: allow list_dir in analyst tools (follows the
list_dir wiring in 968cbfa).
- QA plan doc for the branch sweep.
ContextSummarizedEvent (registered) keyed on stable session sequence, not the
ephemeral pack ordinal, so replay reuses the recorded summary artifact and never
re-calls the LLM (invariants #6/#8). TierContextSummarizer mirrors
JournalCompactionService for freeform context (tool outputs/docs/retrieved text)
the journal path doesn't cover. ContextState + DefaultContextReducer fold the
event; reducer replay-safety unit-tested.
Three defects surfaced by a live headless freestyle_planning run (all made a
long execution workflow fail or stall in ways invisible to a REST operator):
1. retryCount was session-global (DefaultOrchestrationReducer): TransitionExecuted
updated currentStageId but never reset retryCount, while shouldRetry gates on
it per-stage. An early stage that burned its 3 retries starved every later
stage of its own — the next stage's first *retryable* failure went terminal.
Reset retryCount on stage entry; back-edge loops stay bounded by the separate
refinementIterations counter. Regression test added.
2. Freestyle stages ran at StageConfig's 4096-token default (ExecutionPlanCompiler
never set a budget) vs 16384-32768 for static role_pipeline stages. A tool-heavy
stage truncated its context every round, evicting the file_read results it just
gathered, so it re-read forever and never accumulated enough to write. Set 16384.
3. GET /events returned the OLDEST `limit` events (readFrom(0) then take): for a
session past `limit` events the tail — including a pending ApprovalRequested —
was invisible to a headless/REST poller, the exact caller POST /approve serves.
Default now tails; explicit fromSeq still paginates forward.
Two fixes that together let the freestyle analyst/architect stages actually
produce valid artifacts on a local model (verified end-to-end: analyst passed
first try, architect produced a validated plan):
1. Tools-less final emission (SessionOrchestrator): producing the artifact
while tools are in the request is unreliable — llama.cpp's Gemma template
switches to a channel format and leaks <|channel> markers into the text, and
models keep tool-calling until the round cap without ever emitting. After the
tool loop, fire ONE tools-less inference to get clean JSON (also re-enables
the JSON grammar, since grammar+tools is rejected). Gated on the artifact not
already being produced via emit_artifact.
2. GBNF grammar handles arrays/objects (GbnfGrammarConverter): array-typed
properties fell through to a 'string' rule, so the grammar forced a string
where the schema demanded an array — an unwinnable grammar-vs-validator
disagreement. Add generic value/object/array rules and map types correctly.
Note: also needs schema 'items' on array props (runtime schemas updated; mirror
to repo schemas/*.json).
Two robustness fixes for LLM-emitted artifact stages (e.g. freestyle analyst/
architect) that were exhausting their retry budget on weak models:
1. emit_artifact meta-tool: stages with an LLM-emitted slot now expose an
emit_artifact tool whose parameters ARE the artifact's JSON schema, giving
the model a structured tool-call channel to produce the artifact instead of
a free-form final message. Sidesteps llama.cpp's grammar+tools
incompatibility (the artifact can't be grammar-constrained while tools are
present). The call's arguments are captured as the artifact content.
2. Actionable retry feedback: validation rejection now surfaces the concrete
error messages (e.g. 'required property summary missing') as the stage
failure reason instead of a bare 'validation failed', so the retry-feedback
entry tells the model what to fix.
Add CapabilityAwareRoutingStrategy — it hard-filters to providers that declare every
required capability (like FirstAvailable) but ranks the matches by summed required-capability
score; ties keep list order, so it is a strict superset of first-available. The server now
wires it as the routing policy.
The orchestrator augments a stage's required capabilities with ModelCapability.ToolCalling
when the stage grants tools, so tool-heavy stages route to the best tool-calling model rather
than whichever healthy provider comes first.
Scores come from each provider's declared capabilities(); observed per-model reliability
(GET /metrics/tool-reliability) can feed in later as an additional weight.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Enable autonomous QA through a remote OpenAI-compatible provider (NVIDIA NIM)
and harden the tool/approval path so unattended multi-stage runs complete.
- inference: add openai_compat provider (Bearer chat-completions for NIM/OpenAI),
dispatched by provider type "nim"/"openai"; key via api_key/api_key_env.
- server: bind configured [server] host/port instead of a hardcoded 8080;
POST /sessions accepts an optional `intent` (WS parity) for intent-driven workflows.
- kernel: thread the bound operator profile's approval_mode into per-tool gating so
auto/yolo enable unattended approval (engine still consulted; policy/plane-2 BLOCK
stays terminal); on a recoverable tool failure feed the tool's arg-schema back into
context so the model self-corrects instead of repeating a malformed call.
- tools: split deletion out of file_write into a separate, explicitly-named file_delete
tool — a model can no longer delete a file by getting a write-mode parameter wrong.
- server: add GET /metrics/tool-reliability — per-model tool-call validity from the
event log (measurement groundwork for capability-aware routing).
- docs: update AGENTS.md across kernel, tools, server, inference.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
When a write is blocked for being outside the claimed task's
affected_paths, the implementer can propose widening scope via a new
T2 propose_scope tool. The operator decides (invariant #4 — the agent
can't self-grant). Approve unions the proposed paths into the task's
affected_paths (existing TaskAffectedPathsSetEvent); activeScope
re-derives the wider manifest from events so the same write then passes.
Reject blocks the task (via TaskClaimCoordinator.blockActiveTask, hooked
on both approval-denial branches, scoped to propose_scope) so the loop
advances. No new event types.
A stage flagged claimTask gets the kernel (not the LLM) to claim the
next ready task — or re-surface the one already claimed on a review
bounce — and inject its TaskContextAssembler bundle as L0. While a task
is claimed, plane-2's write manifest narrows to the task's affectedPaths.
Project identity is resolved from the origin-session link the task tools
already record (TaskService.projectForSession), so the loop predicate and
claim stage agree regardless of the free-form project key the agent passed.
Claiming flows through a kernel-declared TaskClaimCoordinator implemented
in apps/server (no cross-core import).
New TasksReady condition driven by EvaluationContext.readyTaskCount,
fed at resolve time via a kernel-declared ReadyTaskCounter interface
implemented in apps/server over TaskService (keeps core:kernel
decoupled from core:tasks). Loops a graph while ready>0, exits at 0.
Root Child DOX Index assembled, plus a per-module AGENTS.md across the tree
(core/*, infrastructure/*, apps/*, testing/*, and docs/examples/frontend/etc),
each following the DOX section shape: Purpose, Ownership, Local Contracts,
Work Guidance, Verification, Child DOX Index.
Two orchestrator tool-gating behaviors (both touch SessionOrchestrator):
- Read-only switch: when a tool call is BLOCKed with READ_BEFORE_WRITE, derive a
read-only mode from the event log (block until a later FILE_READ completes) and
withhold FILE_WRITE tools from the next inference turn(s), so a weak model is
pushed down the read-then-write path instead of looping on the blocked write.
- require_task_decompose stage flag (TomlWorkflowLoader) + owesTaskDecompose guard:
a stage so marked cannot stage_complete until a task_decompose/task_create has
succeeded this stage. New task_planning workflow + prompt are the first consumer
(decompose-only: swoop codebase, emit a parent + DEPENDS_ON task graph, stop).
GET /tasks now reports dependency-graph status per task — ready (TODO with no unmet
deps) and blockedBy (ids of unfinished blockers) — resolved via TaskGraph over each
task's FULL project board, so a status filter can't hide a blocker. Default-valued
fields are omitted (encodeDefaults=false); consumers read absent as zero.
The task_decompose approval card now shows a readable graph (epic + numbered children
with their "after" dependency labels) via renderDecomposePreview wired into
computeToolPreview, instead of the truncated raw JSON the generic fallback gave.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The orchestrator stringified every non-primitive tool argument, so a JSON array
arrived as its toString() and listParam (as? List<*>) read it empty — silently
dropping a task's affected_paths/acceptance_criteria. That left WriteScopeRule
(activeTask.scope empty -> no-op) and cite-before-claim with nothing to enforce
for agent-created tasks.
Extract parseToolArguments: a primitive array becomes List<String>; objects and
arrays-of-objects keep their JSON text for tools that re-parse them
(task_decompose). Unit-tested in ParseToolArgumentsTest.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Block writing a file whose on-disk content changed since the session read it — a
concurrent/external edit the agent's view doesn't reflect. file_read records a
content hash on whole-file reads; the orchestrator's completion path now carries
the tool's structuredOutput (it previously dropped it, unlike SandboxedToolExecutor
— the two paths were inconsistent), so SessionContext.readHashes folds it.
StaleWriteRule compares the current hash (WorldProbe.contentHash) against the
read-time hash and BLOCKs a mismatch, telling the agent to re-read first. Files the
session wrote itself are excluded, so its own edits never look stale; partial reads
set no baseline.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Replace the single-purpose sessionReadPaths plumbing with a SessionContext
{reads, writes, activeTask} folded by SessionContextProjection over the session
stream — the read-model every anti-hallucination gate consumes (plane-2 rules via
ToolCallAssessmentInput.session; tool gates via a port, next).
- reads: completions whose recorded capability includes FILE_READ.
- writes: the resolved paths the orchestrator already records on each completion's
receipt.affectedEntities (FileAffectingTool) — no param-parsing reinvention.
- activeTask: null for now; populated once claim records a session fact.
ReadBeforeWriteRule now reads input.session.reads; ReadFilesProjection is retired.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Capability is the canonical signal the plane-2 gates dispatch on, but
ToolInvocationRequestedEvent only carried the tool name — so ReadFilesProjection
classified reads by a hardcoded "file_read" string, bypassing the abstraction and
(worse) re-deriving a semantic fact at replay from whatever the name meant. That
violates the event-sourcing invariant that replay reads recorded facts, not live
state.
Now the tool's requiredCapabilities are recorded on the event at emission, and the
projection classifies by ToolCapability.FILE_READ. To let the lowest module carry
it, ToolCapability moves from core:tools into core:events keeping its package, so
every existing import resolves unchanged and core:tools imports it from below. The
field is defaulted, so pre-existing events deserialize as empty.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
An LLM that edits or overwrites a file it never read is acting on hallucinated
contents. This plane-2 ToolCallRule blocks it: dispatching on FILE_WRITE (covers
file_write and file_edit), a write to a path that exists on disk but was not
file_read earlier this session is BLOCKED. A brand-new file passes — you can't
read what doesn't exist, and creating is legitimate.
- ReadFilesProjection folds the session's file_read events (request + completion,
so a failed read doesn't count) into the set of paths read, mirroring
EgressAllowlistProjection; threaded into ToolCallAssessmentInput.sessionReadPaths
via SessionOrchestrator.resolveSessionReadPaths.
- Read paths and the write target are both resolved through the probe (toRealPath),
so spelling differences and symlinks can't dodge the gate.
- The orchestrator already feeds a block's rationale back as a tool result, so the
agent reads the file and retries; the rejected-event reason now carries that
rationale too (specific audit trail instead of a generic string).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Two reviewer-reliability tracks, both deterministic and unit-verified (no model/network).
§B-§5 — static_check stage seam. The StaticAnalysisRunner + StaticFindingsRecordedEvent +
reviewer-context filter already existed; what was missing was a stage that runs the tool and
emits the event. Added:
- StaticCheckStageExecutor: reads stage metadata (static_tool/static_argv), runs the configured
command via StaticAnalysisRunner, returns a StaticFindingsRecordedEvent. No-op (empty findings)
when no runner is wired or no command is configured — safe to carry unconfigured.
- A deterministic-stage seam in DefaultSessionOrchestrator.enterStage: any stage with
metadata["stage_type"] == "static_check" is run by the executor instead of the LLM subagent,
then advances on its (unconditional) exit edge.
- TomlWorkflowLoader: stage_type/static_tool/static_argv fields → StageConfig.metadata.
- role_pipeline.toml: a static_check stage between implementer and reviewer (no-op until
static_argv + a CommandRunner are set; activation is live-QA-gated, see StaticAnalysisRunner doc).
§B-§6 — critique-outcome producer. The CritiqueFinding type + CriticCalibrationProjection existed
but nothing fed them. Added:
- CritiqueFindingsRecordedEvent (+ CritiqueVerdict): the producing side — a critic's findings +
verdict for one review iteration, carrying modelHash for per-model calibration.
- CritiqueOutcomeCorrelator: pure loop-resolution logic deciding UPHELD (fixed between rounds) /
DISMISSED (persisted into an approved final) / INCONCLUSIVE (open at a non-approved terminal),
per critic (role + modelHash) and per finding id.
- A hook in completeWorkflow/failWorkflow that correlates recorded findings into
CritiqueOutcomeCorrelatedEvents at loop resolution — no-op when none recorded, idempotent.
(LLM-side finding emission stays a separate model-gated activation.)
Tests: StaticCheckStageExecutorTest (4), StaticCheckStageTest integration (1, fake runner →
event + transition), CritiqueOutcomeCorrelatorTest (8), CritiqueCalibrationWiringTest integration
(1, seeded findings → outcomes at completion), updated RolePipelineWorkflowTest. Full suites for
core:events/kernel/critique, infrastructure:workflow, testing:integration green; detekt clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Make the grant system's wider scopes real and add a revoke producer.
Before: grants were loaded per-session (the gate folds only the running
session's own stream), so GrantScope.PROJECT — though declared and matched
in the engine — was dead in practice, and there was no GLOBAL scope or any
way to revoke a grant (ApprovalGrantExpiredEvent had no producer).
- GrantScope: add GLOBAL(toolName); make PROJECT tool-bound. Every
operator-creatable scope is now tool-bound so a grant can never be a
blanket "approve everything" (that's YOLO mode).
- DefaultApprovalEngine.scopeMatches: GLOBAL matches the bound tool in any
context; PROJECT matches when the session's projectId AND tool match.
- Cross-session ledger: PROJECT/GLOBAL grants are appended to a reserved
GRANT_LEDGER_SESSION_ID stream instead of a session's. The approval gate
now unions the ledger's grants with the session's, and derives projectId
from the bound workspace root (ProjectIdentity.of) so PROJECT grants match
later sessions on the same repo. SESSION/STAGE grants still live in (and
die with) their session.
- Revoke: ClientMessage.RevokeGrant appends ApprovalGrantExpiredEvent to the
ledger (reducer already drops it); ListGrants/GrantList expose the active
standing grants for the TUI viewer.
- Drop the server-side T2 grant ceiling per operator request: a grant may
now authorize any tier (incl. destructive T3/T4). Tool-binding is retained
as the remaining guard.
Backend only; the TUI scope picker + grants viewer follow. core:events,
core:approvals, core:kernel, apps:server compile; approvals/events/kernel/
server suites green; new GrantScopeMatchingTest (5) covers GLOBAL/PROJECT
match, project isolation, and the no-cap T4 path.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
On session start, discover CLAUDE.md and AGENTS.md at the bound workspace root and inject
their (concatenated, header-labeled) contents as an L0 system context entry for every stage —
mirroring the .correx/project.toml ProjectProfile path end to end. Recorded as
AgentInstructionsBoundEvent (invariant #9) so replay reads the recorded fact, not the live
file; folded into SessionState.boundAgentInstructions and rendered by buildAgentInstructionsEntry
right after the project profile. AgentInstructionsLoader reads root-only, both files if present.
Bound via ServerModule.bindAgentInstructions alongside bindProjectProfile.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
ToolCallAssessmentInput gains sessionEgressHosts; SessionOrchestrator resolves it at
the assessment build site by folding EgressHostsGrantedEvents through
EgressAllowlistProjection. NetworkHostRule now enforces session-granted hosts (union
with static), replacing the emptySet() TODO — the allowlist built in 027ff1f is live.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- StaticFinding/StaticFindingSeverity + StaticFindingsRecordedEvent (registered + test)
- StaticAnalysisRunner over an injectable CommandRunner; parses compiler/ktlint/detekt
finding formats (lenient) into StaticFindings
- excludeStaticFindingsFromReview pure filter, live-wired into SessionOrchestrator
context assembly so static-tool findings are kept out of the reviewer's context
- static_check pipeline stage left as a documented TODO (needs an event-emitting
deterministic stage-type seam that doesn't exist yet)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- StageCheckpointPassed/FailedEvent (registered + serialization test): per-stage
reconciliation of produced artifacts vs the locked plan's produces-slots
- StageCheckpointReconciler (pure) + tests
- emitStageCheckpoint wired in after verifyProduces, runs regardless of its
pass/fail (verifyProduces still owns the halt); gated on an ExecutionPlanLocked
in the session log so non-plan workflows are unaffected
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- BriefEcho model + BriefEchoMismatchEvent (registered + serialization test)
- BriefEchoExtractor: parses the planner's brief_echo block and the original
brief's referenced-files/symbols/acceptance-criteria dimensions
- BriefEchoComparator: conservative diff — a dropped acceptance criterion or an
invented file path halts; dropping a file/symbol alone does not
- checkBriefEcho gate chained into the post-stage flow after groundBriefReferences;
opt-in via stageConfig.metadata["briefEcho"], fails the stage (retryable) on
divergence with re-grounding feedback
- prompt/workflow activation (planner.md emitting brief_echo; role_pipeline
metadata) left as live-QA follow-up
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Run operator-configured static-analysis commands (compiler/detekt/formatters) as a
deterministic harness step on the producing stage, before the LLM reviewer. A stage
declares `static_analysis = [commands]`; after it produces its artifact, each command
runs in the workspace root and the results are recorded in a StaticAnalysisCompletedEvent
(invariant #9 env observation — recorded live, folded on replay, never re-run). A
non-clean command fails the stage retryably with its output fed back verbatim (§2:
compiler/test output is ground truth), so the implementer fixes it and only static-clean
code ever reaches the reviewer. This makes §5's "reviewer context excludes static findings"
mechanical — the findings are resolved upstream, so the reviewer can't waste inference on
what detekt catches for free; no output-parsing, no context plumbing.
- StaticAnalysisCompletedEvent + StaticAnalysisFinding (core:events) + registration.
- StaticAnalysisRunner seam + ProcessStaticAnalysisRunner (whitespace-split argv, stderr
merged, timeout→nonzero exit) in core:kernel; wired (nullable) into OrchestratorEngines
and the server. Null runner / no workspace root → logged no-op, never blocks the run.
- StageConfig.staticAnalysis + `static_analysis` TOML field; runStaticAnalysis folded into
a runPostStageGates sequence (produces → grounding → echo → static analysis).
- Documented `static_analysis` on the role_pipeline implementer stage (commented; commands
are workspace-specific). The reviewer prompt half shipped in 3467826.
New brief_echo stage before the planner in role_pipeline: the model restates
the analyst brief as a structured artifact, and a deterministic gate
(checkBriefEcho, opt-in via brief_echo=true) diffs the restatement against the
original brief. On divergence — dropped requirements (Jaccard coverage < 0.34)
or hallucinated files — it emits BriefEchoMismatchEvent and fails the stage
retryably, so a misread brief never reaches plan generation.
The diff (BriefEchoDiff) is a pure function of two recorded artifacts, so it is
replay-safe and records no observation; the event fires only on mismatch, for
audit. Symbols are recorded but non-blocking in v1. Mirrors the groundBrief-
References gate. role_pipeline only; freestyle_planning wiring is a follow-up.
Runtime install (like research): copy brief_echo.json + brief_echo.md and the
[[artifacts]] block into ~/.config/correx.
When a stage produces an LLM artifact carrying a non-empty questions
array, the orchestrator parks the run, records ClarificationRequested,
awaits the operator's answers (submitClarification seam), records
ClarificationAnswered, injects the answered Q&A as an L2 USER entry, and
re-runs the same stage so the questions are resolved by the role that
asked them. Runs on a dedicated pendingClarifications map and the
enterStage success-branch loop, so a clarification round never consumes
the failure-retry budget; capped at three rounds. Integration test covers
the happy path and the cap.
Local analysts hallucinate file paths the same as planners. A stage can now
opt in with `ground_references = true`: after it produces its brief, every
workspace-relative file path the brief named is checked for existence, and a
path that doesn't exist fails the stage (retryable) with the misses fed back
("Reference only files you have read") so the model re-grounds.
Grounding probes the filesystem directly rather than the repo map — the map
is recency-ranked and depth/extension-limited, not a complete existence
oracle, so "absent from the map" can't be a hard failure. Existence is a
nondeterministic observation, so the result is recorded as a
BriefGroundingCheckedEvent (invariant #9); replay reads it back and never
re-probes.
- BriefGroundingCheckedEvent + GroundingReference, registered in eventModule
- BriefReferenceExtractor: pure, deterministic pull of relative file-path
references from a brief artifact JSON (needs a '/' and a dotted extension;
skips modules, bare names, URLs, absolute/parent paths — so coarse
"subsystem" mentions the brief is allowed to make aren't false-flagged)
- SessionOrchestrator.groundBriefReferences on the post-verifyProduces path,
gated by stage metadata; uses the existing worldProbe
- TomlWorkflowLoader `ground_references` → metadata; role_pipeline analyst
opts in
Scoped to file references with a '/' + extension (not bare basenames, which
a depth-limited probe can't resolve), so a flagged miss is high-confidence.
The symbol-grounding half of §3 is left for a real symbol index.
Scope creep — an implementer writing files it never declared — is the
dominant local-model failure that compiler/tests don't catch. A stage can
now declare a `writes` manifest (workspace-relative globs); a FILE_WRITE
that resolves inside the workspace but outside the declared set is raised
as PATH_OUTSIDE_MANIFEST → BLOCK, so the model gets the violation as tool
feedback and retries within scope (the §2 bounded-retry signal).
Implemented as a plane-2 rule beside PathContainmentRule (same effect-based
dispatch on FILE_WRITE, same symlink-safe WorldProbe resolution, same
recorded observations so replay reads them back — invariant #9). An empty
manifest is unrestricted; out-of-workspace targets stay PathContainmentRule's
concern, so the two rules don't double-flag.
- StageConfig.writeManifest + TomlWorkflowLoader `writes` parsing
- ToolCallAssessmentInput.writeManifest, threaded from the active stage via
SessionOrchestrator.runPlane2Assessment
- new ManifestContainmentRule, registered in the server assessor
- shared candidatePathStrings extracted so both path rules judge the same
target set
- role_pipeline.toml documents the option on the implementer stage
The dedicated ManifestViolationEvent the spec names is not added: the
violation is already recorded replayably in ToolCallAssessedEvent (issue +
observations + BLOCK) and surfaced in the TUI rationale band. A first-class
event is worth adding only once reconciliation consumes it.
Layer 1 of tui-requirements §5: a no-LLM analyzer that shell-lexes a
proposed command, classifies binaries against a known-command table, and
mechanically flags the constructs that widen blast radius — sudo/doas,
recursive delete, pipe-to-shell (skipping wrappers so `| sudo sh` counts),
redirects outside the workspace, network binaries, base64/eval, and
unparseable input (itself a red flag). Pure + replay-stable: same preview
+ workspaceRoot → same analysis, no environment reads.
The approval band renders command approvals as a card — worst-first flag
chips over the reconstructed command line, each token colored by severity
(T0–T4 ramp). ^x opens a scrollable fullscreen view so a long command is
never truncated. Diff and plain-preview paths are unchanged.
Server: computeToolPreview now renders the shell tool's argv as a full,
shell-quoted command line instead of the 200-char JSON-args fallback,
honouring §3's "full untruncated command". The TUI parser accepts both
the rendered line and the older `{"argv":[...]}` shape for replay.
Layer 2 (model annotation) deliberately deferred per the spec.
ExecutionPlanCompiler treated the plan's single 'produces' string as a
registry kind id, while the architect prompt described it as a unique
artifact id referenced by needs/edges — the model followed the prompt,
invented descriptive ids (files_created), and every freestyle plan
failed to compile. Collapsing both onto one string is also unsound:
two stages producing the same kind would collide on slot name and make
artifact_validated edge conditions ambiguous.
Plan stages now carry an optional 'kind' selecting the registered kind
(mirroring TOML's produces = { name, kind }); 'produces' stays the
unique slot name. Missing 'kind' falls back to the old produces-as-kind
reading so existing plans still compile. Vocabulary entry, architect
prompt, and execution_plan schemas updated to match.
LlamaCppTokenizer posted {"tokens":[text]} but llama.cpp's /tokenize expects
{"content":text}; the server silently answered {"tokens":[]}, so countTokens
returned 0 for every string. Every context entry carried tokenEstimate=0 and
the whole budget/trim pipeline was blind — live QA saw a 34k-token prompt
sail through a 16384 stage budget (three raw file_read results of plan docs),
crater t/s, degrade output, and fail the stage on validation 3x.
Also guard estimateTokens: a zero count for non-blank content falls back to
the chars/4 heuristic so a lying tokenizer can never blind budgeting again.
Three gaps found by live QA (kill between planning-complete and plan lock):
- launchSessionResumeWithRehydrate never handed off to FreestyleDriver, so a
resumed freestyle session stranded at planning-complete with no plan lock
and no phase 2; now locks+runs when planning completed and no
ExecutionPlanLockedEvent exists yet.
- resume route 400'd on phase-2 sessions (workflowId freestyle-<sid> is
compiled, never registered); now recompiles the locked plan after
rehydrating the artifact cache.
- orchestrator.resume threw when currentStageId was terminal ('done') or
unknown to the graph; now returns WorkflowResult.Completed.
kami