Every other top-level content folder is snake_case without a 'awesome_'
prefix (starter_ai_agents, advanced_ai_agents, always_on_agents...);
this one now matches. All internal references updated: main README,
category README, both skill READMEs, CI workflow paths, evals tooling.
Git tracked the moves as renames, so per-file history is preserved.
Also:
- Model name corrected: 'Gemini Flash' -> 'Gemini 3.5 Flash' (SKILL.md
and README table) — it was already correct everywhere else in the
skill, this was the one straggler
- Main README skills blurb rephrased: 'Browse all skills' folded into
the intro sentence as a link, agent-count claim softened to 'Claude
Code, Codex, Cursor, and other coding agents' instead of a number
All four CI-equivalent gates re-verified against the new path: lint
strict (both skills), security scan, trigger & routing, deterministic
evals — all pass.
Known consequence: this changes the install path for project-graveyard,
already merged to main and already installed by at least one user via
the old awesome_agent_skills/ URL. The old npx skills add command stops
resolving once this merges; anyone who has it installed keeps their
local copy (skills CLI doesn't phone home), but the previously-shared
install link needs reposting with the new agent_skills/ path.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Tier 2 (trigger & routing) is now automated in CI instead of a manual
protocol: run_trigger_evals.py checks that positive prompts clear the
near-miss negatives on description vocabulary alone, and — once the
catalog has 2+ skills — that positives rank their own skill first and
no two descriptions near-collide. Reasoning-triggered cases (necromancer
mode) are marked lexical:false and covered by tier 3.
Tier 3 (behavioral) adopts skill-creator's evals.json schema verbatim,
so Anthropic's run_eval.py, benchmarking, and eval viewer work against
our files unmodified. project-graveyard's four behavioral cases written
in that schema.
The evals README tier table now uses Addy's names for Addy's tiers, and
marks the two tiers that are ours because our skills ship executable
code and his don't (security scan, deterministic script evals).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
One job meant one check. Now every PR touching awesome_agent_skills/
gets the full gate the skills were built under:
- deterministic evals: every evals/*/test_*.py (synthetic fixtures)
- skill lint (strict): every */SKILL.md against the spec — frontmatter
rules, name==dir, unfilled placeholders, text-only prompt dumps
- security scan: install lures, undeclared network, credential access,
obfuscated payloads across all skills
skill_lint.py and skill_scanner.py are vendored into
awesome_agent_skills/evals/tools/ as repo-side tooling. All three pass
locally on the current tree.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Scans your machine for dead side projects, autopsies why each one died
from its git history (deploy fear, payments wall, killed-by-newer-
project kill-chain, boilerplate death), shows your personal death
patterns, and resurrects the one with a pulse. The agent interviews
you on ambiguous deaths, world-checks what got easier since, executes
the resurrection plan, and tracks relapses on every resurrection it
prescribes.
Python stdlib, offline, read-only. The installable folder contains
only what runs at runtime; the 16-assertion executable eval lives in
awesome_agent_skills/evals/ — run it from the clone before installing.
One-command install via the skills CLI (70+ agents).
Also: self-improving-agent-skills drops its toy example files — its
examples endpoint now surfaces sibling skills from this repo, so the
app demos on real skills.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Shubhamsaboo