context-window-to-skill
Skill153 starsChanged 44 days agoMITView on GitHub
Takes the active conversation as reference to understand how a skill can be created, with all the lessons learned from the user's needs in the conversation.
In this repository41 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- .github/copilot-instructions.mdCopilot instructions
- broken-access-controlSkill
- cdn-peekSkill
- http-request-smugglingSkill
- jwt-attacksSkill
- ssrf-oobSkill
- webapp-probeSkill
- blockchain-forensicsSkill
- foundry-pocSkill
- safe-huntSkill
- smart-contract-auditSkill
- cloud-bucket-bruteSkill
- infrastructure-auditSkill
- dylib-hijack-scanSkill
- endpoint-threat-huntSkill
- parse-securitySkill
- agentic-seoSkill
- agent-onboardingSkill
- git-commitSkill
- keen-eyesSkill
- no-aispeakSkill
- secure-1pass-accessSkill
- self-improveSkill
- shadow-partnerSkill
- training-guideSkill
- variant-tableSkill
- auditor-quizSkill
- audit-scopeSkill
- evidence-captureSkill
- external-enumerationSkill
- gdocs-audit-reportSkill
- hackerone-reportSkill
- nuclei-suggestSkill
- pre-bountySkill
- sandboxed-audit-runnerSkill
- tiny-auditorSkill
- vm-labSkill
- llm-fuzz-toolsSkill
- mcp-server-pentestSkill
- prompt-injection-auditSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 14
Create a skill and place it on the ~/.claude/skills/ directory.
≈ “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
Use Case:
User had finished working on his feature or task successfully, but had to tweak lots of stuff in the behavior of the agent prior to completion - we now want to create a skill that in the next time will make this task a breeze
Steps:
- Analyze conversation and understand fully what the user tried to achieve
- Analyze where were the pitfalls, user frustrations, and what pre-information would have made the results quicker to get with the most quality
- Print to the user 10 (or more) top items you think are the core points to base the new skill upon
- Consult with the user if questions still remain
- Create a skill and place it on the ~/.claude/skills/ directory. comply to agent skill format and https://code.claude.com/docs/en/skills
Results:
- Working new skill, that next time would do the job much better with all the tweaks needed to perfect it based on learned conversation and observation
- New skill should be reusable (e.g. think fixes also for future usecases other than the single one the user encountered)
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.