Forked from Carlo1911/skill-evolution main. Adds env-var override so a single
installed pipeline can target per-profile skills trees (default ~/.hermes/skills,
krystie ~/.hermes/profiles/krystie/skills, sami-story ~/.hermes/profiles/sami-story/skills)
without monkeypatching HOME or the SKILLS_DIR module constant.
- scripts/skill_index.py: add resolve_skills_dir() + SKILLS_DIR_ENV_VAR, wire into
scan_skills() with explicit-arg > env-var > default resolution order. Read at
call time (matches fetch_sessions.get_state_db_path() convention).
- tests/test_skill_index.py: 2 new tests covering nonblank env var and blank
fallback (the latter monkeypatches SKILLS_DIR + invokes scan_skills() with no
args so the test actually exercises the fallback expression, not just the
explicit arg path).
- 49/49 conformance tests pass.
- Codex grade: A (verified 2026-08-05).
Co-Authored-By: Claude Opus 4 <noreply@anthropic.com>
Standalone Python stdlib pipeline that reads an agent's past sessions,
compares them against installed skills, and generates structured
improvement proposals gated by an evaluation framework before anything
mutates. Host-agnostic via HostAdapter (Hermes, Claude Code).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Krystie