Skip to content

feat: close evidence instruments and add bounded recovery - #11

Merged
shrec merged 1 commit into
mainfrom
agent/strix-recovery-findings
Aug 5, 2026
Merged

feat: close evidence instruments and add bounded recovery#11
shrec merged 1 commit into
mainfrom
agent/strix-recovery-findings

Conversation

@shrec

@shrec shrec commented Aug 5, 2026

Copy link
Copy Markdown
Owner

Summary

  • explicitly close and regression-verify the 16-item evidence/economics instrument set
  • add bounded, non-executing MCP recovery guidance for hallucinated tool names
  • add line-movement-stable root-cause fingerprints and duplicate summaries to static scanner evidence
  • document the source-pinned Strix capability review without importing upstream code or parallel authorities
  • bump release projections to 0.8.99

Evidence

  • Python: 2027 passed, 27 skipped
  • evidence-instrument closure suite: 175 passed
  • extension suite: passed, 29 files
  • Ruff: passed
  • Source Graph, semantic-edit, and context-envelope benchmark checkers: passed
  • wheel, sdist, Twine, and VSIX packaging: passed

Truth boundaries

  • implementation closure does not fabricate risk-precision or A/B populations
  • unknown-tool suggestions never alias or execute a tool
  • root-cause grouping never upgrades static candidates to runtime-validated findings

Workspace hygiene

The two pre-existing B829 eval files remain uncommitted and are excluded from this PR.

@shrec
shrec marked this pull request as ready for review August 5, 2026 15:52
Copilot AI lite review requested due to automatic review settings August 5, 2026 15:52
@shrec
shrec merged commit f85367a into main Aug 5, 2026
10 checks passed

Copilot AI left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

This PR bumps AIWorkHub to 0.8.99 and tightens “evidence-first” behavior by (1) making MCP unknown-tool failures recoverable with bounded, deterministic guidance, and (2) adding stable root-cause fingerprints plus duplicate summaries to static Known Bug Scanner evidence (including SARIF output), alongside roadmap/changelog documentation.

Changes:

  • Add unknown_tool_message() and wire it into both stdio MCP dispatchers so unknown tools return bounded JSON recovery guidance and suggestions (no auto-aliasing/execution).
  • Extend Known Bug Scanner output with root_cause_fingerprint, dedupe_summary, and SARIF partialFingerprints.aiworkhubRootCauseFingerprint.
  • Version bump and documentation updates (Strix capability review, roadmap and changelog).

Reviewed changes

Copilot reviewed 15 out of 16 changed files in this pull request and generated 2 comments.

Show a summary per file
File Description
vscode-extension/package.json Bumps VS Code extension version to 0.8.99.
vscode-extension/package-lock.json Aligns lockfile package version to 0.8.99.
vscode-extension/extension.js Updates expected MCP package version constant to 0.8.99.
src/aiworkhub/_version.py Bumps Python package version to 0.8.99.
src/aiworkhub/tool_recovery.py Introduces bounded unknown-tool recovery message generator.
src/aiworkhub/stdio_fastmcp.py Emits bounded recovery JSON for unknown tools in dependency-free stdio server.
src/aiworkhub/server.py Emits bounded recovery JSON for unknown tools in fallback stdio dispatcher.
src/aiworkhub/known_bug_scanner.py Adds stable root-cause fingerprints, dedupe summary, and SARIF partial fingerprint.
src/aiworkhub/quality_evidence.py Includes dedupe summary in known-bug evidence check summary.
tests/test_tool_recovery.py Tests determinism, boundedness, and redaction behavior of unknown-tool guidance.
tests/test_stdio_fastmcp_schema.py Ensures unknown-tool errors don’t terminate subsequent stdio requests.
tests/test_known_bug_scanner.py Validates fingerprint stability across line movement and dedupe summary correctness.
tests/test_aiworkhub_embedded_mcp_runtime_b844_v1.py Validates bounded unknown-tool behavior in embedded stdio protocol dispatch.
docs/STRIX_CAPABILITY_REVIEW.md Documents source-pinned Strix capability review and adopted/non-adopted ideas.
docs/PRODUCT_ROADMAP.md Notes evidence-instrument closure and documents Strix-derived refinements.
CHANGELOG.md Adds 0.8.99 release notes describing the new recovery/fingerprint behavior.
Files not reviewed (1)
  • vscode-extension/package-lock.json: Generated file

Comment on lines +17 to +20
def _bounded_name(value: Any) -> str:
text = str(value or "")[:MAX_REQUESTED_CHARS]
match = _SAFE_NAME.fullmatch(text)
return text if match else "<invalid>"
Comment on lines +32 to +37
suggestions = difflib.get_close_matches(
name,
candidates,
n=MAX_SUGGESTIONS,
cutoff=0.45,
)
@shrec
shrec deleted the agent/strix-recovery-findings branch August 5, 2026 15:57
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants