agentsclimarketplace

Contract check

Skill flurdy/agent-skills/skills/contract-check

Shared AI agent skills

Install
npx -y skills add flurdy/agent-skills --skill contract-check

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 6 stars6 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Audit health of contract tests across services — staleness, sync gaps, uncommitted pacts, missing tests.

SKILL.md

9.6 KB, as published. Nobody here has run it

Contract Check — Pact Health Auditor

Audits the health of consumer-driven contract tests across all letterbox services. Surfaces staleness, sync gaps, uncommitted pacts, missing tests, documentation drift, and disabled tests.

This is a read-only audit — it does not run tests or modify files. Use /contract-test to execute contract test workflows.

When to Use

  • Starting a work session to check contract health
  • After modifying connectors or API endpoints
  • Before a release to verify all contracts are synced and committed
  • Periodically to catch drift and gaps

Usage

/contract-check              # Full audit (all checks)
/contract-check status       # Summary dashboard only
/contract-check stale        # Consumer output newer than provider copy
/contract-check uncommitted  # Pact files not committed in service repos
/contract-check sync-gaps    # Trace each intended edge: consumer test → built pact → synced to provider
/contract-check coverage     # Provider CI verifies each synced pact? (catches commented-out/disabled consumers)
/contract-check missing      # Connectors with no consumer tests
/contract-check docs         # Documentation drift in pact-workflow.md
/contract-check disabled     # Contract tests excluded from default builds
/contract-check <service>    # All checks scoped to one service

Setup

On first run, ensure the script symlink exists:

ln -sfn "$SKILLS_DIR/contract-check/scripts/contract-check.sh" ./scripts/contract-check
chmod +x ./scripts/contract-check

Where $SKILLS_DIR resolves to ${CLAUDE_HOME:-$HOME/.claude}/skills.

Instructions

Step 0: Ensure Setup

Check that ./scripts/contract-check exists and is executable. If not, create the symlink as shown in Setup above.

Step 1: Parse Subcommand

Parse the user's argument:

  • (none) or full → Run all checks (mechanical + semantic)
  • status → Run all checks but only show the summary table
  • stale → Run ./scripts/contract-check stale
  • uncommitted → Run ./scripts/contract-check uncommitted
  • sync-gaps → Run ./scripts/contract-check sync-gaps
  • coverage → Run ./scripts/contract-check coverage (CI verifies each synced pact?)
  • missing → Run semantic Missing Consumer Tests check (Step 3)
  • docs → Run semantic Documentation Drift check (Step 4)
  • disabled → Run semantic Disabled Tests check (Step 5). coverage is the mechanical companion: it catches commented-out PACTCONSUMER entries in a provider's CI verify job.
  • <service> → Run all checks scoped to that service
  • matrix → Run ./scripts/contract-check matrix

Step 1b: Normalize Before Checking (recommended)

Before running the uncommitted check (or all/full), suggest running make normalize-pacts first. Pact libraries regenerate random UUIDs, dates, and strings on every test run, causing noisy git diffs that aren't real contract changes. The normalizer replaces these with deterministic placeholders.

Note: normalize-pacts.py currently only normalizes values referenced by generators.body metadata. UUIDs in providerStates.params, request.path, request.headers, and response.headers are NOT yet normalized — those may still show as noise in uncommitted diffs. Flag these as noise in the report when the diff is UUID-only.

Step 2: Mechanical Checks (via script)

Run the appropriate script subcommand:

./scripts/contract-check all      # or stale, uncommitted, sync-gaps, coverage, matrix

Parse the script output. Each line starts with a status keyword:

  • OK — no action needed
  • STALE — consumer pact is newer than provider copy
  • DIFFERS — same age but content differs
  • MISSING_PROVIDER — provider does not have this pact file
  • UNCOMMITTED — pact file has uncommitted changes
  • NOT_BUILT — (sync-gaps) consumer test exists but produced no pact in target/ (tests not run)
  • NOT_SYNCED — (sync-gaps) pact is built but not copied into the provider (run sync-pacts.sh)
  • GAP — (coverage) provider CI does not verify all its synced consumer pacts (style=enum with commented-out consumers, or style=none with no verify job)
  • CLEAN — no issues found
  • NO_DATA — no consumer pact files found
  • SUMMARY — counts for the check

Step 3: Missing Consumer Tests (LLM-driven)

For each consumer service, compare connectors against consumer test files:

  1. Find connectors — Glob for connector files:

    • Play services (admin, hosted, registration, profile, web): <service>/app/connectors/*Connector.scala
    • Scala 3 services (digest, patrol, reconciler, dispatch, etc.): <service>/src/main/scala/**/connectors/*Connector.scala

  2. Find consumer tests — Glob for consumer test files:

    • Play services: <service>/test/contract/*Consumer*.scala
    • Scala 3 services: <service>/src/test/scala/**/contract/*Consumer*.scala

  3. Extract provider name from connector filename: strip Connector suffix, lowercase. E.g., AccountConnector.scala → account.

  4. Exclude external connectors that are not internal services:

    • Braintree, Stripe, SES, Spam, Web, Api (base class), ConnectorConfiguration, ConnectorModule

  5. Cross-reference: For each connector, check if a matching consumer test exists. Report connectors with no consumer test.

Output format:

### Missing Consumer Tests
| Service | Connector | Provider | Has Consumer Test |
|---------|-----------|----------|-------------------|
| reconciler | EventConnector | event | NO |
| reconciler | MessageQueueConnector | messagequeue | NO |

Step 4: Documentation Drift (LLM-driven)

  1. Read docs/pact-workflow.md
  2. Run ./scripts/contract-check matrix to get actual relationships
  3. Compare the documented consumer→provider relationships against the matrix
  4. Flag:
    • Consumers listed in docs but not in matrix (or vice versa)
    • Providers listed in docs but not in matrix (or vice versa)
    • Stale TODOs or notes that have been resolved
    • Missing entries for newer services (digest, patrol, reconciler, profile)

Step 5: Disabled Tests (LLM-driven)

  1. Search for contract test exclusions in build.sbt files:

    • Pattern: -l tags.ContractTest or --exclude-tags=ContractTest
    • Check both active and commented-out exclusions

  2. For services with active exclusions, check if the Makefile test-contract target overrides it (e.g., with set Test/testOptions := Nil)

  3. Report:

    • Services where sbt test silently skips contract tests
    • Whether the Makefile target correctly overrides the exclusion
    • Services lacking a test-contract Makefile target entirely

Step 6: Format Report

Combine all findings into a health report:

# Contract Health Report

## Summary
| Check            | Status | Details                        |
|------------------|--------|--------------------------------|
| Staleness        | PASS/WARN/FAIL | X stale / Y total       |
| Uncommitted      | PASS/WARN | X files across Y services     |
| Sync coverage    | PASS/FAIL | X/Y pairs in sync-pacts.sh   |
| Verify coverage  | PASS/FAIL | X providers verify all synced pacts |
| Missing tests    | PASS/WARN | X connectors without tests    |
| Documentation    | PASS/FAIL | X items out of date           |
| Disabled tests   | PASS/INFO | X services exclude by default |

## Staleness
[details from script output]

## Uncommitted Pact Files
[details from script output]

## Sync Coverage Gaps
[details from script output]

## Missing Consumer Tests
[details from LLM analysis]

## Documentation Drift
[details from LLM analysis]

## Disabled Tests
[details from LLM analysis]

## Recommended Actions (priority order)
1. [most impactful fix first]
2. ...

Status thresholds:

  • PASS: No issues found
  • INFO: Informational, no action required
  • WARN: Issues found but not blocking
  • FAIL: Significant gaps that should be addressed

Step 7: Offer Remediation

After presenting the report, ask the user:

Would you like me to create beads for the issues found? I can triage them as P3 tasks with appropriate service labels.

If the user agrees, use /triage to create beads for actionable findings. Group related findings into single beads where appropriate (e.g., "Add digest, patrol, reconciler to sync-pacts.sh" as one bead rather than three).

Scoped Checks (single service)

When invoked with a service name (/contract-check dispatch):

  1. Run mechanical checks and filter output to only show lines involving that service (as consumer or provider)
  2. Run semantic checks scoped to that service only
  3. Present a focused report for just that service

Error Handling

  • Script not found: Create the symlink (see Setup)
  • mgit not found: Error — must be run from within the letterbox project root
  • No consumer pact files: Report that consumer tests haven't been run; suggest make test-contract in consumer services
  • Service directory missing: Skip gracefully, note in output

Known Provider Pact Directories

Most providers use test/resources/pacts/. Exceptions:

  • event: src/test/resources/pacts/
  • membership: src/test/resources/pacts/

Known External Connectors (exclude from missing-test checks)

Braintree, Stripe, SES, SpamConfiguration, WebConfiguration, ApiConnector (abstract base), ConnectorConfiguration, ConnectorModule

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.