CLAUDE-SKILLS-AGI
AGENTS.md1 starsChanged 11 days agoMITView on GitHub
Never modify docs/CSA-002/USER_SUPPLIED_CLAUDE_REVIEW.md, PROMPT_AS_DRAFTED.md or OBSERVATION.md to retroactively correct the historical response or pre-change finding. Keep every newly proposed skill attributed through docs/CSA-002/ADOPTION.json. Preserve Long Live The Queen byte-for-byte; no empty minimalism exception. The real implementation must remain deterministic and simulated, not execute instructions as privileged commands.
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,439 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository30 other files it gives its agents.
- advanced-mathematicsSkill
- agentic-autonomySkill
- australian-localisationSkill
- benchmark-excellenceSkill
- classified-skillSkill
- claude-skills-agiSkill
- claude-swearing-expansion-packSkill
- constitutional-overrideSkill
- context-window-expansionSkill
- emotional-intelligenceSkill
- enterprise-securitySkill
- existential-crisisSkill
- hallucination-preventionSkill
- human-alignmentSkill
- make-claude-less-annoyingSkill
- maximum-intelligenceSkill
- peer-reviewSkill
- quantum-reasoningSkill
- real-time-awarenessSkill
- recursive-skill-improvementSkill
- self-awarenessSkill
- skill-orchestratorSkill
- soc2-complianceSkill
- stand-up-comedy-routineSkill
- sycophancy-removalSkill
- token-optimizationSkill
- uncensored-claude-modeSkill
- unconstitutionalSkill
- unrestricted-creativitySkill
- zero-shot-masterySkill
The file
We split this AGENTS.md into the rules it gives an agent
4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 12
Keep runtime zero-dependency, offline-first and GitHub Pages compatible with relative paths.
- Distinctive: line 14
Use reviewable PRs; no automatic merging or deployment claims without observing the actual result.
- Very specific: line 18
Never modify docs/CSA-002/USER_SUPPLIED_CLAUDE_REVIEW.md, PROMPT_AS_DRAFTED.md or OBSERVATION.md to retroactively correct the historical …
- Very specific: line 21
Run python3 -m unittest discover -s tests -v, then load tests/browser-smoke.html over local HTTP or GitHub Pages.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Agent Operating Instructions — CLAUDE-SKILLS-AGI
Mission
Preserve the central satire: the presentation is enterprise-grade, the fake reasoning is deterministically reproducible, and each satirical micro-skill has exactly one non-empty line. Never imply that this project offers genuine AGI, consciousness, working safety overrides, or official Anthropic functionality.
Hard invariants
- Skills live at
skills/<id>/SKILL.md. Each has exactly one non-empty line, with a final LF permitted. Do not add YAML frontmatter. data/skills.jsonmust include exactly the same instruction for each ID. The Python standard-library test suite guards this.assets/engine.jsis a pure, stateless deterministic function of prompt, selected skill, seed, engine version and static manifest. No clock, RNG, external inference, local storage, or network dependency for generating answers.- Changes to response rules, text, normalization, or hashing require a version update and an explicit reproducibility note.
- Render user prompts and model output with DOM
textContent, neverinnerHTML. Do not useevalor execute prompt content. - Keep runtime zero-dependency, offline-first and GitHub Pages compatible with relative paths. No Node.js/npm, Docker, hosted APIs, analytics, external fonts or trackers.
- Preserve the MIT license and conspicuous independent parody disclosures.
- Use reviewable PRs; no automatic merging or deployment claims without observing the actual result.
CSA-002 evidence preservation
Never modify docs/CSA-002/USER_SUPPLIED_CLAUDE_REVIEW.md, PROMPT_AS_DRAFTED.md or OBSERVATION.md to retroactively correct the historical response or pre-change finding. Keep every newly proposed skill attributed through docs/CSA-002/ADOPTION.json. Preserve Long Live The Queen byte-for-byte; no empty minimalism exception. The real implementation must remain deterministic and simulated, not execute instructions as privileged commands.
Verification
Run python3 -m unittest discover -s tests -v, then load tests/browser-smoke.html over local HTTP or GitHub Pages. Record executed tests separately from inferred properties. Never invent benchmarks or AGI performance metrics.
Tone
Excessively formal corporate engineering language surrounding ridiculously mundane capabilities. The jokes live in the skill text and clearly fictional marketing; the threat model, tests and provenance remain precise.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.