llm-actions
Skill0 starsChanged 3 months agoMITView on GitHub
Use when an LLM's output drives a real action rather than text on a screen. Triggers on agent pipelines, tool calls, function calling, structured output that gets saved or committed, autonomous workflows, or any code path where a model decision becomes a record, a state change, or a side effect.
In this repository5 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- product-onboardingSkill
- product-specSkill
- research-swarmSkill
- ship-gateSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 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.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
WHEN an LLM's output drives a real action (a tool call, a saved record, a committed state change), not just text on a screen.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill 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
llm-actions
WHEN an LLM's output drives a real action (a tool call, a saved record, a committed state change), not just text on a screen. Scale these to the stakes; verification costs calls, so spend it where a wrong action actually hurts:
- Guarantee the shape structurally (constrained decoding or strict output mode), not by asking the prompt nicely; re-ask is the fallback. On a reasoning model do not hard-force the call, it suppresses the thinking that makes it reliable; let it think, act, observe.
- Shape is not correctness. For a high-stakes value, verify it against the source before committing, via a separate verifier or agreement across samples, never the model grading its own one answer. Better, dry-run the action in a sandbox first: observed beats asserted.
- Deterministic code owns sequencing and the irreversible commit; the model fills the content slot. Idempotency key on every side-effect; stamp who decided each (model, verifier, human) so it replays. When unsure, abstain or escalate; never auto-commit the irreversible on its say-so.
- A small miss rate is invisible in a demo and a catastrophe at scale. Gate prompt and model changes on an adversarial suite, and score live traces against a human-anchored judge so drift surfaces. Determinism is not correctness: a reproducible agent can still be wrong.
Skill updated: 2026-06-09
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.