- Home
- Skills
- Different Ai
- Agent Bank
- Skill Reinforcement
skill-reinforcement_skill
- TypeScript
186
GitHub Stars
1
Bundled Files
2 months ago
Catalog Refreshed
4 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill different-ai/agent-bank --skill skill-reinforcement- SKILL.md7.7 KB
Overview
This skill automatically captures learnings and anti-patterns after any skill run and immediately improves the skill corpus. It prevents knowledge loss between sessions by analyzing outcomes, extracting shortcuts, documenting failures, and updating skill files with actionable edits. The goal is continuous, low-friction improvement after every use.
How this skill works
After a skill completes, the skill inspects the execution context, outcome, token usage, and time taken. It asks focused questions (what failed, what was faster, what assumption broke) and classifies discoveries into categories like shortcuts, failure modes, better patterns, anti-patterns, and environment quirks. Findings are formatted with templates and appended to the relevant skill’s documentation, then validated for clarity and non-redundancy. Periodic reviews surface high-impact areas across many skills.
When to use it
- Automatically after any skill completes (success, partial, or failure)
- When a workflow succeeds or fails in a notable or repeatable way
- Whenever a new shortcut or token-saving pattern is discovered
- If commands fail and you identify a reliable fix
- When API behavior differs from docs or assumptions prove wrong
Best practices
- Capture context immediately: skill name, task, outcome, token cost, and time
- Use concise, actionable entries with exact commands or code snippets
- Classify each learning into categories (shortcut, failure, anti-pattern, etc.)
- Validate each update for formatting, specificity, and non-redundancy
- Cross-pollinate learnings to related skills when applicable
Example use cases
- Found OTP visible in email preview: add a token-saving shortcut to the OTP flow
- Encountered session persistence in staging: add a check-first pattern to login workflows
- Fixed a failing curl invocation: document the exact command fix under Common Issues
- Observed repetitive manual steps: convert repeated actions into a new skill/tool
- Noticed inconsistent API responses: update prerequisites and add an environment quirk note
FAQ
Apply updates immediately after a meaningful discovery; avoid batching notes for much later.
What counts as an actionable learning?
Anything with a clear fix, exact command, or measurable benefit (token/time reduction) that others can follow.