feat(skills): catalog with signed manifest installs#1812
Merged
Conversation
Promote the MCP catalog browse / list / search / install / upgrade / info / status surface to skill packs. Source variants (github, git, npm, file, directory) resolve through the existing plugin_installer; catalog manifests carry an Ed25519 signature that the install verifies against the catalog's signer_pubkey. Every install / upgrade appends a skill.catalog.install event to the HMAC-chained audit log with (manifest_url, manifest_sha256, manifest_signer_pubkey, install_id, prev_chain_digest); replaying the chain pulls the identical sha and refuses installation if the upstream sha drifted. skills.lock is extended with [[catalog]] rows and [[lineage_receipt]] rows so two parallel worktrees launched from the same chain head observe identical skill versions, and an upgrade in one worktree produces a deterministic adopt/pin decision in the other. The existing lineage-v1 gate is extended with check_skill_lockfile to reject PRs whose lockfile references a manifest sha not present in the chain's known-good set. Catalog cache lives under .sdd/skills_catalog/ with revalidation honouring BERNSTEIN_SKILLS_CATALOG_TTL. Closes #1796
![sourcery-ai[bot]](https://avatars.githubusercontent.com/in/48477?s=60&v=4){kind=small}
There was a problem hiding this comment.
Sorry @chernistry, you have reached your weekly rate limit of 2500000 diff characters.
Please try again later or upgrade to continue using Sourcery
|
Warning Rate limit exceeded
You’ve run out of usage credits. Purchase more in the billing tab. ⌛ How to resolve this issue?After the wait time has elapsed, a review can be triggered using the We recommend that you space out your commits to avoid hitting the rate limit. 🚦 How do rate limits work?CodeRabbit enforces hourly rate limits for each developer per organization. Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout. Please see our FAQ for further information. ℹ️ Review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: ASSERTIVE Plan: Pro Run ID: ⛔ Files ignored due to path filters (1)
📒 Files selected for processing (16)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
Contributor
Sonar insights (advisory, no merge-block)Snapshot of
Run This comment is a soft signal. The Sonar scan runs on push to |
Contributor
Review-bot acknowledgement summary
All must-address findings are resolved or acknowledged. |
Contributor
|
bernstein doctor observe for PR #1812 ( sonar -- WARN (project bernstein)
code-scanning -- WARN (27 open alert(s))
Skipped backends (credentials not configured)
See docs/observability/unified-doctor.md for backend setup notes. |
Contributor
Mutation gate (fixed critical paths)
Surviving mutations (top 20)
|
Contributor
Contract drift detected - proposed patchInline autofix push failed ( Three contract tests act as drift detectors against the public CLI / API surface:
One or more failed on this PR. Files changed: How to applyEither run the regen script locally: uv run python scripts/regen_contract_drift.py --fixture all
git add -A && git commit -m "chore(ci): regenerate contract drift allow-lists"
git pushOr apply the patch directly: gh pr checkout 1812
git apply <<'PATCH'
diff --git a/tests/unit/test_readme_api_coverage.py b/tests/unit/test_readme_api_coverage.py
index fa6b08a6..90b3d71d 100644
--- a/tests/unit/test_readme_api_coverage.py
+++ b/tests/unit/test_readme_api_coverage.py
@@ -239,6 +239,8 @@ DOCUMENTED_COMMANDS: frozenset[str] = frozenset(
"desktop-register",
# Bot-added: drift autofix (regen_contract_drift.py)
"supervisor",
+ # Bot-added: drift autofix (regen_contract_drift.py)
+ "schedule",
}
)
PATCH
git add -A && git commit -m "chore(ci): regenerate contract drift allow-lists"
git pushFull diffdiff --git a/tests/unit/test_readme_api_coverage.py b/tests/unit/test_readme_api_coverage.py
index fa6b08a6..90b3d71d 100644
--- a/tests/unit/test_readme_api_coverage.py
+++ b/tests/unit/test_readme_api_coverage.py
@@ -239,6 +239,8 @@ DOCUMENTED_COMMANDS: frozenset[str] = frozenset(
"desktop-register",
# Bot-added: drift autofix (regen_contract_drift.py)
"supervisor",
+ # Bot-added: drift autofix (regen_contract_drift.py)
+ "schedule",
}
)Source CI run: https://github.com/sipyourdrink-ltd/bernstein/actions/runs/26251528553 Refs #1273. |
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
| from rich.table import Table | ||
|
|
||
| service = _build_service(scope) | ||
| rows = service.list_installed() |
| def info_cmd(entry_id: str, refresh: bool, scope: str) -> None: | ||
| """Show full info for a single catalog entry.""" | ||
| service = _build_service(scope) | ||
| entry = service.info(entry_id, force_refresh=refresh) |
| def sync_cmd(scope: str) -> None: | ||
| """Detect lockfile vs on-disk drift.""" | ||
| service = _build_service(scope) | ||
| drift = service.sync() |
| def status_cmd(scope: str) -> None: | ||
| """Show cache + lockfile state for ``skills catalog``.""" | ||
| service = _build_service(scope) | ||
| status = service.status() |
| return GateResult(ok=False, failures=["skill catalog lockfile module is missing"]) | ||
|
|
||
| state = read_state(lockfile_path) | ||
| failures: list[str] = [] |
| swaps the upstream while keeping the URL is caught. | ||
| """ | ||
| canonical = json.dumps(payload, sort_keys=True, separators=(",", ":")).encode("utf-8") | ||
| hasher = hashlib.sha256() |
| swaps the upstream while keeping the URL is caught. | ||
| """ | ||
| canonical = json.dumps(payload, sort_keys=True, separators=(",", ":")).encode("utf-8") | ||
| hasher = hashlib.sha256() |
| plugin_installer: InstallerCallable | None = None, | ||
| preloaded_catalog: SkillCatalog | None = None, | ||
| ) -> None: | ||
| if fetcher is None and preloaded_catalog is None: |
| ) -> UpgradeOutcome: | ||
| """Upgrade a single installed catalog entry.""" | ||
| state = read_state(self.lockfile_path) | ||
| prior = state.find_catalog(entry_id) |
| raise SkillCatalogError(f"{entry_id!r} is not installed") | ||
|
|
||
| catalog = self.browse(force_refresh=force_refresh) | ||
| upstream = catalog.find(entry_id) |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Summary
Promote the MCP catalog browse / list / search / install / upgrade / info / status surface to skill packs. Source variants (
github,git,npm,file,directory) resolve through the existingplugin_installer; manifests carry an Ed25519 signature that install refuses unless--allow-unverifiedis set. Each install appends askill.catalog.installaudit-chain event with(manifest_url, manifest_sha256, manifest_signer_pubkey, install_id, prev_chain_digest); replaying the chain pulls the identical sha and refuses if upstream drifted.skills.lockis extended with[[catalog]]rows and[[lineage_receipt]]rows so two parallel worktrees launched from the same chain head observe identical skill versions; an upgrade in one produces a deterministic adopt/pin decision in the other. The existing lineage-v1 gate is extended (not replaced) withcheck_skill_lockfileto reject PRs whose lockfile references a manifest sha not present in the chain's known-good set. Catalog cache lives under.sdd/skills_catalog/with revalidation honouringBERNSTEIN_SKILLS_CATALOG_TTL.Closes #1796
Files touched
src/bernstein/cli/commands/skills_catalog_cmd.py(new): CLI surfacesrc/bernstein/cli/main.py: wirebernstein skills catalogsubgroupsrc/bernstein/core/skills/catalog/__init__.py(new): public re-exportssrc/bernstein/core/skills/catalog/manifest.py(new): strict schema + validatorsrc/bernstein/core/skills/catalog/signature.py(new): Ed25519 sign/verifysrc/bernstein/core/skills/catalog/fetcher.py(new): HTTP fetcher + cachesrc/bernstein/core/skills/catalog/audit.py(new): audit chain integrationsrc/bernstein/core/skills/catalog/installer.py(new): wrapsplugin_installersrc/bernstein/core/skills/catalog/lockfile.py(new): atomic lockfile + receiptssrc/bernstein/core/skills/catalog/service.py(new): high-level servicesrc/bernstein/core/lineage/gate.py: additivecheck_skill_lockfiletests/unit/core/skills/test_catalog.py(new): 25 unit teststests/integration/test_skill_catalog_cross_worktree.py(new): 3 integration testsdocs/operations/skills-catalog.md(new): operator docsCHANGELOG.md: Unreleased entryAcceptance criteria
bernstein skills catalogexposes browse, list, search, install, upgrade<id>/--all, info, status modelled onbernstein mcp catalog(manifest_url, manifest_sha256, manifest_signer_pubkey, install_id, prev_chain_digest); replay refuses on upstream sha drift--allow-unverified.sdd/skills_catalog/withBERNSTEIN_SKILLS_CATALOG_TTLenv overrideplugin_installerand writes into.bernstein/skills/<name>/skills.lockupdates atomically;bernstein skills catalog syncdetects drift; two parallel worktrees from the same chain head observe identical skill versions; upgrade in wt-a produces a lineage receipt that wt-b uses to decide adopt or pinTest plan
uv run pytest tests/unit/core/skills/test_catalog.py tests/integration/test_skill_catalog_cross_worktree.py(28 tests pass)uv run pytest tests/unit/skills/(250 existing skill tests pass)uv run pytest tests/unit/lineage/test_gate.py(23 lineage-gate tests pass)uv run pytest tests/unit -k "skill or catalog"(603 tests pass)uv run ruff checkclean on all new/touched filesbernstein skills catalog --helpexposes all subcommands