@firstpick/pi-skill-deep-research

A Pi skill for high-stakes or complex research needing multi-source evidence, scientific/technical fact-checking, decision traces, or rigorous verification. Runs deterministic two-phase research with schema/policy validation.

What it does

• Adds the deep-research skill to Pi's skill library.
• Guides agents to invoke the skill for high-stakes or complex research needing multi-source evidence, scientific/technical fact-checking, decision traces, or rigorous verification. Runs deterministic two-phase research with schema/policy validation.
• Bundles skills/deep-research/SKILL.md plus any supporting references, scripts, tests, fixtures, or assets used by the skill.

Install

pi install npm:@firstpick/pi-skill-deep-research

Configuration

No required configuration.

Expected usage structure

The skill bundles deterministic research policy, schema, state, tests, and a runner script. Agents should resolve these paths relative to the installed skill directory.

Bundled layout:

skills/deep-research/
  SKILL.md
  policy.json
  output-schema.json
  state.json
  scripts/
    run_deep_research.py
  tests/
    test_determinism.py

When following the skill manually, run from skills/deep-research/ or use absolute paths:

cd /path/to/installed/package/skills/deep-research
python3 ./scripts/run_deep_research.py \
  --policy ./policy.json \
  --schema ./output-schema.json \
  --state ./state.json \
  --input /path/to/input.json \
  --output /path/to/output.json

The packaged state.json is intended as skill-local run state. If you want project-local history instead, copy it into the project and pass that copied path with --state.

Commands

None.

Tools

None.

Example view

User: Review this change for the concerns covered by `deep-research`.
Agent: Invokes the `deep-research` skill, follows its workflow, and reports the result.