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>
This commit is contained in:
2026-08-27 11:57:49 +04:00
parent 519290368f
commit 6a8a7b31c1
36 changed files with 1232 additions and 79 deletions
+2
View File
@@ -16,6 +16,8 @@ Adapter for `core:tools`. Depends on `core:tools`, `core:events`, `core:approval
- Web search and web fetch results are environment observations; they must be recorded as events by callers to preserve replay determinism (invariant #9).
- `ToolConfig` is the only configuration surface; pass via `InfrastructureModule.createToolExecutor()`.
- `buildTools()` extension on `ToolConfig` assembles the full tool list; add new tools there, not in the registry directly.
- `file_copy` copies one existing file to another path (`{source, dest}`) so static/binary assets never pass through the model's token stream. Same jail, anchor and `fileWrite.enabled` toggle as `file_write`; `dest` is the mutated target (`ParamRole.PATH`, the only affected path), `source` is read-only (`ParamRole.SOURCE_PATH`) and may sit under an operator-granted out-of-workspace path exactly as a `file_read` may. One regular file per call: no recursion, no globs.
- Every path parameter resolves through `ToolPath.resolve` (`core:tools`), which expands a leading `~` and anchors relatives on the bound workspace's working dir. Do not re-implement path resolution in a tool.
- Filesystem mutation is split by intent: `file_write` only writes (`{path, content}`), `file_edit` edits, and `file_delete` only deletes (`{path}`) — deletion is a separately-named capability so a model can never delete by getting a write-mode parameter wrong. `file_delete` shares `file_write`'s path jail and `fileWrite.enabled` toggle and carries `ToolCapability.FILE_WRITE`.
- `list_dir` is shallow by default, but collapses a non-symlink single-child directory chain (bounded depth) to the first branch point and explains that expansion in its output; recursive listings retain normal tree traversal.