FCIS / Effect Skills for Pi

Eight composable Agent Skills for Functional Core / Imperative Shell architecture, with Effect 4 guidance activated only when the project evidence proves the exact installed Effect version.

Skills

Planning

• fp-domain-modeling — clarify domain terms, invariants, and transitions.
• fp-grill-with-docs — interview through design decisions and persist the glossary, decision ledger, and selective ADRs.
• fp-codebase-design — map functional cores, capability ports, adapters, package boundaries, and roots.
• fp-to-spec — synthesize approved evidence into an implementation contract.

Execution

• fp-tdd — deliver one core-first vertical slice at a time.
• fp-implement — implement an approved spec through validated tracer slices.
• fp-diagnosing-bugs — reproduce, minimize, instrument, and fix bugs by boundary seam.
• fp-code-review — review Standards, Spec, and FCIS/Effect concerns independently.

The recommended workflow is fp-grill-with-docs → fp-codebase-design → fp-to-spec → fp-implement with fp-tdd for each slice, followed by fp-code-review. Domain modeling and bug diagnosis can also be used independently.

Install with Pi

Install the npm package globally so the skills are available across your projects:

pi install npm:fp-skills@1.0.1

Or install it only for the current project:

pi install --local npm:fp-skills@1.0.1

The npm package includes the pi-package keyword used for Pi package gallery discovery. You can also install the same release directly from its Git tag:

pi install git:github.com/emre-yildiz-dev/fp-skills@v1.0.1

Project package declarations load only after Pi project trust is granted. Review the package and its skill instructions before trusting a project. Check configured packages with pi list; update the npm package with pi update npm:fp-skills, or the Git package with pi update git:github.com/emre-yildiz-dev/fp-skills; remove the source you installed with the corresponding pi remove command.

To update to a newer release, change the pinned npm version or Git tag in the install command. Pinned versions remain fixed until you install a newer version.

You can force a skill in Pi with /skill:fp-codebase-design or /skill:fp-tdd. The package contains standard .agents/skills directories and can also be copied into a compatible project's .agents/skills/ directory for other Agent Skills implementations.

Validation

From the repository root, run the deterministic release checks:

python3 -m unittest discover -s .agents/skills/fp/scripts/tests -p 'test_*.py' -v
python3 .agents/skills/fp/scripts/validate_suite.py --suite-root .agents/skills/fp --through 2
uv lock --check
uv run --frozen ruff check .agents/skills/fp/scripts

The repository also includes quality cases, trigger cases, fixtures, and an Effect-backend scenario runbook. Model/provider-dependent evaluations require a configured Pi provider and are not claimed as passing by the deterministic CI workflow; see evals/fp-suite/effect-backend-scenarios.md for the bounded template-validation scenarios.

Distribution

The skills are published as the npm package fp-skills and are eligible for Pi package gallery discovery through the pi-package keyword. The same source is also available from GitHub as a tagged Pi Git package.

License and attribution

The upstream workflow source is adapted from mattpocock/skills at commit c55ee46073ed923f86ce59a5eb3b6d895095d1b7. Its MIT notice is retained in .agents/skills/fp/UPSTREAM-LICENSE.md. The FCIS/Effect guidance and Pi integration are maintained in this repository and may diverge from upstream.