You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Refresh and compact codebase-memory-mcp for 0.12.1 🤖🤖🤖 - #6410
The published skill targets 0.11.0 and repeats its workflow in a separate reference section. This refreshes it for the 0.12.1 MCP contract and reduces SKILL.md from 16,172 to 6,928 bytes.
The compact workflow covers index freshness and documentation coverage, grouped qualified names, independent search streams, architecture jobs and generation changes, MENTIONS, and infrastructure traces. It retains explicit mutation authorization, rooted source fallback, container source mode, relative coverage paths, and snapshot-identity checks. The generated skill catalog follows the shorter trigger description. The description instructs the agent to skip simple value-only lookups; activation is not strictly enforced.
For explicitly approved ADR edits, it also documents a tested limitation: set_sections replaces unfenced level-two headings and appends unknown names. Nested or fenced text requires a full-document update with readback, preserving unrelated text.
npm run build: passes; the generated skill catalog is up to date.
Repository line-ending normalizer and git diff --check: pass. The local macOS run uses a BSD-sed compatibility adapter.
Contract check against native MCP 0.12.1: 17 tools and 36 instantiated scenario inputs; YAML checks and 44 local runner self-tests pass.
Native ADR controls reproduce the wrong nested-section append, verify exact full-document replacement, and verify ordinary level-two section replacement preserves the remainder.
Three seeded Codex runs of the preceding ADR revision and three of the local variant: 6/6 exact ADR edits, full reads and readbacks, zero erroneous intermediate section writes. Each uses an isolated native MCP and read-only shell, allowing only the requested fixture ADR mutation. All temporary daemons and indexes are cleaned up.
The preceding compact ADR revision achieved source-checked correct outcomes in all 12 scenarios across three repeats (36/36), matching the saved long-guide baseline's 36/36. The latest two-line clarification was checked separately with three read-only value lookups and three explicitly authorized index operations: 6/6 correct outcomes. These task-outcome counts include documented corrections of grader false positives: raw scores were 27/36 for each 36-run series and 5/6 for the final series. Corrections were checked against source and tool traces; raw results are retained. Read-only median input tokens changed from 98,791 to 66,387, while median command count stayed at 3; this does not establish general efficiency or speedup. These tests used isolated native MCP 0.12.1 fixtures and Codex gpt-6.1-sol / xhigh / priority.
The host-specific evaluation harness is not included. Direct GitHub Copilot execution remains untested.
Pull Request Checklist
Read and followed CONTRIBUTING.md.
Paid-service submission guidance is not applicable to this existing local, MIT-licensed MCP workflow.
Update to an existing skill in the correct directory; file names and format validated.
Ran the repository build and regenerated the catalog.
Targets main.
Tested this revision with GitHub Copilot.
Type of Contribution
Update to existing skill.
By submitting this pull request, I confirm that my contribution abides by the Code of Conduct and will be licensed under the MIT License.
Risk tier:merge-risk:medium — New or substantially changed resource without privileged execution Required to merge: passing submission-gate checks plus 1 approval from reviewers with write access.
✓ [spec-compliance] All 1 skill(s) are spec-compliant.
ℹ️
✓ spec-compliance: All spec checks passed.
ℹ️
✓ [valid-refs] All file references across 1 skill(s) are valid.
ℹ️
✓ valid-refs: All file references resolve to existing files within the skill directory.
ℹ️
1 skill(s) linted, 1 passed
Full linter output
### Linting skills/codebase-memory-mcp
✅ codebase-memory-mcp (2/2 checks passed)
✓ [spec-compliance] All 1 skill(s) are spec-compliant.
✓ spec-compliance: All spec checks passed.
✓ [valid-refs] All file references across 1 skill(s) are valid.
✓ valid-refs: All file references resolve to existing files within the skill directory.
1 skill(s) linted, 1 passed
The reason will be displayed to describe this comment to others. Learn more.
🔵 Needs a closer look
The interacting pagination, freshness, async-indexing, and ADR-write contracts require maintainer verification because the 0.12.1 implementation and host-specific evaluation harness are unavailable in this checkout.
No open issues 🎉
What changed in this PR
Updates the existing Codebase Memory MCP skill from 0.11.0 to 0.12.1, consolidating its workflow and reference material into a shorter guide.
Changes:
Documents updated pagination, index freshness, architecture jobs, documentation links, and infrastructure tracing.
Retains mutation authorization and rooted source fallbacks, and clarifies safe ADR editing.
Synchronizes the generated skill catalog with the shorter activation description.
File
Description
skills/codebase-memory-mcp/SKILL.md
Compacts the guide and updates tool usage and safety guidance for 0.12.1.
docs/README.skills.md
Updates the catalog entry to match the revised skill description.
The reason will be displayed to describe this comment to others. Learn more.
🔵 Needs a closer look
The revision lacks a captured GitHub Copilot validation run on a real repository, leaving its client-specific behavior unverified.
No open issues 🎉
1 previously missed issue
In code that hasn't changed since last review
Add GitHub Copilot validation evidence
skills/codebase-memory-mcp/SKILL.md:14
The PR reports native MCP and Codex fixture validation but explicitly leaves this revision untested with GitHub Copilot, so its client-specific behavior remains unverified. Please add a captured GitHub Copilot run on a real repository to the PR description, identifying the selected project/index, the evidence inspected, and a source-checked conclusion.
Review effort: Balanced (adaptive)
This branch has not been deployed
No deployments
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
The published skill targets 0.11.0 and repeats its workflow in a separate reference section. This refreshes it for the 0.12.1 MCP contract and reduces SKILL.md from 16,172 to 6,928 bytes.
The compact workflow covers index freshness and documentation coverage, grouped qualified names, independent search streams, architecture jobs and generation changes, MENTIONS, and infrastructure traces. It retains explicit mutation authorization, rooted source fallback, container source mode, relative coverage paths, and snapshot-identity checks. The generated skill catalog follows the shorter trigger description. The description instructs the agent to skip simple value-only lookups; activation is not strictly enforced.
For explicitly approved ADR edits, it also documents a tested limitation:
set_sectionsreplaces unfenced level-two headings and appends unknown names. Nested or fenced text requires a full-document update with readback, preserving unrelated text.Follows #3430, which updated the skill to 0.11.0.
Validation
npm run skill:validate: all 463 skills valid.npm run build: passes; the generated skill catalog is up to date.git diff --check: pass. The local macOS run uses a BSD-sed compatibility adapter.The preceding compact ADR revision achieved source-checked correct outcomes in all 12 scenarios across three repeats (36/36), matching the saved long-guide baseline's 36/36. The latest two-line clarification was checked separately with three read-only value lookups and three explicitly authorized index operations: 6/6 correct outcomes. These task-outcome counts include documented corrections of grader false positives: raw scores were 27/36 for each 36-run series and 5/6 for the final series. Corrections were checked against source and tool traces; raw results are retained. Read-only median input tokens changed from 98,791 to 66,387, while median command count stayed at 3; this does not establish general efficiency or speedup. These tests used isolated native MCP 0.12.1 fixtures and Codex
gpt-6.1-sol/xhigh/priority.The host-specific evaluation harness is not included. Direct GitHub Copilot execution remains untested.
Pull Request Checklist
main.Type of Contribution
By submitting this pull request, I confirm that my contribution abides by the Code of Conduct and will be licensed under the MIT License.