Agent skill
manual-regression-api
Manual backend regression via curl-like HTTP checks against the docker-compose stack. Required for API use-case changes (label: curl).
Install this agent skill to your Project
npx add-skill https://github.com/frostaura/ai.toolkit.gaia/tree/main/.github/skills/manual-regression-api
SKILL.md
Manual Regression (API via docker-compose + curl)
When to use
Use when:
- Orchestrator flags use-case change affecting an HTTP API, OR
- Drift fix (“code wins”) changes behavior/use-cases, OR
- Automated integration tests are not yet feasible but validation is required.
Manual regression is required for API use-case changes.
Inputs
- Use-case docs:
/docs/use-cases/UC-###-*.md - Running docker-compose stack (required)
- API base URL/port
- Auth credentials/tokens (if needed)
Outputs
- Manual verification that UC acceptance criteria hold at the HTTP boundary
- MCP proof label recorded:
curl - Any automated integration tests added recorded under
tests_added[](if created)
Rules
- Must run against docker-compose stack (docker-first prerequisite).
- Do not paste response bodies/logs in chat.
- If secrets/credentials are missing: add MCP blockers/questions and keep completion gated.
Step 1 — Start the stack
Preferred:
make upConfirm:- API reachable (health endpoint if available)
- dependencies healthy
If not reachable:
- fix compose/runtime first (blocking)
Step 2 — Create a UC checklist (compact)
For each affected UC:
- list 3–8 checks max:
- happy path request/response
- auth failure or validation failure (if relevant)
- key state change (if UC implies persistence)
Step 3 — Execute curl-style checks
Perform minimal calls that prove acceptance criteria:
- Use the correct headers/auth
- Validate status codes
- Validate presence of key fields (not full payload diffs)
If you discover undocumented behavior:
- trigger spec-consistency work (drift is blocking)
Step 4 — Handle missing info (needs input)
If you cannot complete checks due to missing:
- base URL/ports
- credentials
- seed data requirements Call MCP “needs input” and continue with other tasks.
Step 5 — Completion proof (MCP)
When marking task done:
manual_regression[]must include:curltests_added[]includes any new integration test files createdchanged_files[]includes any config/doc changes
Suggested required_gates[]
- For API use-case change:
manual-regression(and typicallyintegration) plus baselinelint/build/ci
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
stack-default-dotnet-api
Opinionated baseline for .NET HTTP APIs: analyzers/formatting, unit+integration strategy, docker-compose, Makefile targets, and CI wiring.
tasking-and-proof
How the orchestrator must create/manage MCP tasks (todo/doing/done), set required_gates, handle blockers/questions, and record low-context proof for completion.
integration-testing-http
Run curl-style integration checks against the docker-compose stack for HTTP APIs. Required for use-case changes.
spec-consistency
Prevent and repair drift between `/docs` (source of truth), code, tests, CI, and runtime artifacts. Use before marking work done.
gaia-process
End-to-end Gaia SDLC workflow (Repo Explorer → drift/CI fixes → task graph → gated delivery → QA veto → MCP proof). Use for any work in a repo.
stack-default-web-ts
Opinionated baseline for JS/TS web repos: lint/format, tests, Playwright, docker (if HTTP API), Makefile targets, and CI wiring.
Didn't find tool you were looking for?