feat: keyless /pandavas skill mode over a deterministic verb spine - #2
Merged
Conversation
A second way to run pandavas, with no API key: a /pandavas command for Claude
Code and Cursor where the host harness's model plays the LLM agents (research/
worker/judge) while LLM-free CLI verbs own every deterministic gate. Strictly
additive - the standalone `python -m pandavas run` path and the six
deterministic modules are unchanged.
- src/pandavas/spine.py: baseline, resolve-brief, run-tests, judge-gate,
decide, restore, commit, diff, apply-edits. Reuses run's logic via
nodes.make_bhima_test/_better/_restore_snapshot; state in a gitignored
.pandavas/ dir; ASCII JSON stdout; distinct exit codes (0/10/20/30).
- cli.py: register + dispatch the verb subparsers (run path untouched).
- retrieval.py + .gitignore: ignore the .pandavas/ working dir.
- Packaging: commands/, skills/, agents/ (fresh-context judge subagent),
scripts/install-{claude,cursor}.{sh,ps1}, plugin manifests, .gitattributes.
- docs/SKILL_MODE.md, README skill-mode section, CLAUDE.md scope note.
- tests: test_spine.py (per-verb) + test_skill_e2e.py (keyless e2e).
Suite green: 120 passed, 1 skipped.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Adds a keyless
/pandavasskill mode alongside the existing standalone runner. The host harness's model (Claude Code / Cursor) plays the LLM agents (research/worker/judge); LLM-free CLI verbs own every deterministic gate. Two modes share one deterministic core.Why
Run pandavas with no API key, as a one-line
/pandavas . <task>from any repo's terminal, without giving up the owned orchestration.How
src/pandavas/spine.py- verbs:baseline,resolve-brief,run-tests,judge-gate,decide,restore,commit,diff,apply-edits. Reusesrun's logic (nodes.make_bhima_test/_better/_restore_snapshot); state in a gitignored.pandavas/; ASCII JSON; exit codes 0/10/20/30.cli.pyregisters + dispatches the verbs (therunpath is untouched).commands/,skills/,agents/(fresh-context judge subagent),scripts/install-*, plugin manifests.docs/SKILL_MODE.md, README section, CLAUDE.md note.Honest caveat
Skill mode can't structurally force the gates the way
rundoes - the model must call the verbs and obey their exit codes.python -m pandavas runremains the hard-determinism path. (Cursor's judge degrades to self-review; Claude Code uses a real subagent.)Tests
Strictly additive.
python -m pytest-> 120 passed, 1 skipped (+27 new:test_spine.pyper-verb,test_skill_e2e.pykeyless end-to-end).