Files

T

claude 6a8a7b31c1 fix(events,tools,toolintent): failure attribution, one path normalization rule, file_copy (#713 )

Three generic harness fixes from the web-ui postmortem dataset. Nothing here keys
on a language, framework, build tool or task type.

1. Failure attribution. WorkflowFailedEvent carries one primary FailureAttribution
   (AGENT | HARNESS | WORKFLOW | ENVIRONMENT | PROVIDER | OPERATOR | UNKNOWN),
   defaulted to UNKNOWN so pre-field events replay unchanged. FailureAttributor is
   the deterministic reason->layer mapping, used both at emission and when
   classifying history, so the baseline and the live metric are one measurement.
   Emission sites set it: failWorkflow derives from the reason unless the caller
   knows the layer, cancellation is OPERATOR, the server catch-all falls back to
   HARNESS, a grounding-rejected plan is AGENT. Multi-cause chains stay on
   FailureTicketOpened — no second causal structure.

   GET /metrics/failure-attribution (FailureAttributionInspectionService, mirroring
   ToolReliabilityInspectionService) reports counts, share, UNKNOWN share, the
   preserved reasons and the ticket categories from the same sessions. Read-only:
   historical events are classified at READ time and reported as `inferred`, never
   written back over an append-only log.

   Baseline over the local log, 122 terminal failures: AGENT 51 (41.8%),
   OPERATOR 28 (23.0%), WORKFLOW 19 (15.6%), PROVIDER 15 (12.3%), HARNESS 6 (4.9%),
   ENVIRONMENT 3 (2.5%), UNKNOWN 0.

2. The `~` guard bug. ToolPath is now the ONE canonical normalization rule
   (expand a leading `~`/`~/`, keep absolutes, anchor relatives on the session
   working dir). Every filesystem tool, all six plane-2 path rules and the approval
   preview resolve through it, so policy and existence checks inspect the path the
   tool will operate on. `~/.gradle/init.d/offline.gradle` used to resolve to
   `<workspace>/~/.gradle/...`: reported non-existent AND in-workspace, so the
   reference gate called a real file a hallucination and the out-of-workspace prompt
   never fired. Containment and external-read approval behaviour are unchanged —
   the expanded path is simply outside the workspace, where it always belonged.

3. file_copy (#713). A first-class tool with the writer's jail, tier, receipt,
   replay and CAS pre/post images; static and binary assets no longer move through
   the model's token stream. Needed one generic split: ParamRole.SOURCE_PATH marks a
   path a call reads FROM, so containment gates judge both params while write-target
   gates (read-before-write, stale-write, write scope, write manifest) judge the
   mutated one. ReadBeforeWriteRule exempts any call declaring a SOURCE_PATH: its
   content comes from disk, not from memory, and requiring a read of a binary is
   unsatisfiable. Existing tools declare no SOURCE_PATH, so their behaviour is
   byte-identical.

Tests: ToolPathTest (9), FailureAttributionTest (10), PathNormalizationRuleTest (6),
FileCopyToolTest (10), plus a home-relative FileReadTool read. ./gradlew check green.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

2026-08-27 11:57:49 +04:00

2.8 KiB

Raw Blame History

core/tools — AGENTS.md

Purpose

Tool contract, registry, executor, and receipt tracking: defines the Tool interface, manages tool registration, executes tools within their declared tier, compresses tool output, and projects tool invocation history via the standard event-sourcing stack.

Ownership

CORREX kernel team.

Local Contracts

Tool — primary interface all tools implement. Must declare Tier (from core:events).
ToolRegistry — holds all registered tools for a session; looked up by name.
ToolExecutor — executes a Tool given a ToolRequest, records ToolEvents in core:events.
ToolResult — sealed result type: success with output, or failure with error.
ToolState rebuilt from ToolEvents via DefaultToolReducer + ToolProjector.
DefaultToolRepository wraps EventReplayer<ToolState>.
ToolInvocationRecord / ToolInvocationStatus / ToolCallAssessmentRecord — history types.
FileMutationRecord — records file-affecting side effects.
FileAffectingTool — extended Tool interface for tools that write files; must declare affectedPaths.
OutputCompressionSpec / ToolOutputCompressor / DeclarativeCompressor — compress large tool outputs to fit token budgets (Hard Invariant #6: compressed output is informational; original events are preserved).
ParamRole — annotates tool parameter semantic roles: PATH (the path a call acts on / mutates), SOURCE_PATH (a path it only reads from while acting on another target, e.g. file_copy), EXEC_COMMAND, NETWORK_TARGET. Plane-2 gates dispatch on these, never on tool names.
ToolPath — the ONE canonical path normalization rule: expands a leading ~/~/ to the user home, keeps absolute paths, anchors relative paths on the session's working dir (never the JVM cwd). Every filesystem tool and every plane-2 path rule must resolve through it so policy checks and execution act on the same path. ~other/… is deliberately NOT expanded.
ValidationResult — result from tool-level parameter validation (pre-execution check).
Hard Invariant #5: all tool side effects are captured in events. Silent execution is not allowed.

Work Guidance

Follow the standard Events→State→Reducer→Projector→Repository pattern (see core/AGENTS.md).
DefaultToolReducer only does state.copy(...).
ToolExecutor must record a ToolExecutedEvent (or equivalent) before returning, even on failure.
Output compression must not discard the original event — only produce a derived summary alongside it.
New tools implement Tool (or FileAffectingTool) and register in ToolRegistry. Do not instantiate tools inside ToolExecutor.

Verification

./gradlew :core:tools:test --rerun-tasks

Child DOX Index

No child AGENTS.md (leaf module).

2.8 KiB Raw Blame History