- Python 100%
| Filename | Latest commit message | Latest commit date |
|---|---|---|
Juan M. Bello-Rivas
6810dddd77
Validate the skill with skills-ref and security-scan it with
SkillSpector. Neither tool ships a pre-commit hook manifest, so both run
as local hooks with pre-managed dependency environments.
The wrappers work around two CLI quirks: skills-ref rejects the repo
root because Path(".").name is empty, and SkillSpector scans the raw
working tree, which would flag gitignored __pycache__ as shipped
bytecode.
|
||
| assets | ||
| references | ||
| scripts | ||
| tests | ||
| .gitignore | ||
| .pre-commit-config.yaml | ||
| README.md | ||
| SKILL.md | ||
Research Claim Ledger
An Agent Skills package for maintaining an evidence-gated ledger of research claims. It is inspired by the control-plane model in ResearchLoop and extends it with first-class action-speaks and audited Lean 4 evidence.
The skill is host-neutral. Its root SKILL.md works with GitHub Copilot, Pi, and other agents that
implement the Agent Skills standard. The bundled CLI uses only Python's standard library.
Install
From this local checkout:
DO_NOT_TRACK=1 npx skills add ~/sources/research-claim-ledger \
--global \
--agent github-copilot \
--agent pi
After publishing the repository:
npx skills add OWNER/research-claim-ledger \
--global \
--agent github-copilot \
--agent pi
Both agents also discover the shared ~/.agents/skills/ directory, so a direct symlink is a
dependency-free alternative:
mkdir -p ~/.agents/skills
ln -s ~/sources/research-claim-ledger ~/.agents/skills/research-claim-ledger
Pi users can explicitly load the skill with:
/skill:research-claim-ledger
Copilot selects it from the description in SKILL.md when a request concerns research claims,
evidence, formal verification, or manuscript auditing.
Quick start
python3 scripts/claim_ledger.py init /path/to/project \
--title "Study title" \
--question "Does method A improve outcome M over baseline B?" \
--falsification-boundary "A prespecified comparison showing no improvement."
python3 scripts/claim_ledger.py propose /path/to/project \
--rq RQ-0001 \
--class empirical \
--statement "Under condition X, method A improves M over baseline B." \
--scope "Condition X and prespecified metric M" \
--falsification "The prespecified comparison shows no improvement."
python3 scripts/claim_ledger.py validate /path/to/project
Run python3 scripts/claim_ledger.py --help for all commands. The CLI stores project state under
.claim-ledger/; it does not alter source code or research artifacts.
Development
python3 -m unittest discover -s tests -v
python3 scripts/claim_ledger.py --help
No package installation is required.