Agent Self-Improvement Log
Paste the following prompt into your AI chat to install this skill:
Please follow https://skillhub.cn/install/skillhub.md and install @user_75908c09/scsst.
About this skill
Problem
Coding agents often make mistakes, get corrected, or hit tool and API failures, but those signals usually vanish with the session. scsst writes those events into project-local Markdown files so they can be reviewed, linked, and reused instead of living only in chat history.
How It Works
It initialises a .learnings/ directory in the workspace and maintains three files:
ERRORS.mdfor command failures, exceptions, and external tool issuesLEARNINGS.mdfor corrections, knowledge gaps, and best practicesFEATURE_REQUESTS.mdfor requested capabilities
Each entry gets a stable ID such as ERR-20250115-A3F and can be connected to related entries with See Also. When a learning recurs across tasks, the skill distils it into a short rule and promotes it to CLAUDE.md, AGENTS.md, or .github/copilot-instructions.md. It also supports ingesting Pattern-Key candidates from simplify-and-harden, using Recurrence-Count and a 30-day window to decide whether a pattern should enter long-lived context.
Boundaries
It is useful for engineering memory, local or team-shared, but not for storing sensitive data. Do not record tokens, private keys, environment variables, or full configuration files unless the user explicitly asks. Cross-session sharing and hook reminders should be used only in trusted environments.
Use Cases
- Log repeated npm/pnpm install failures in Claude Code to `ERRORS.md`, then promote the fix rule to `CLAUDE.md`.
- Record user corrections about API fields and client regeneration in `LEARNINGS.md` and link similar entries.
- Ingest recurring `simplify-and-harden` patterns, dedupe by `Pattern-Key`, and decide promotion to agent context.
- Write tool gotchas into `TOOLS.md` and workflow improvements into `AGENTS.md` in an OpenClaw workspace.
Best For
- Engineers using Claude Code or Codex who need to turn session corrections into durable project memory.
- Agent developers maintaining OpenClaw workspaces who want behavioral patterns and tool gotchas in long-lived context.
- Team owners managing Copilot or AGENTS.md context who need recurring errors distilled into reusable rules.
Related Skills
Turn vague requests into structured prompts with role, task, context, and constraints, outputting only usable prompts without executing the task.
A skill for game NPC and enemy AI that separates decision, steering, and pathfinding, covering FSMs, behavior trees, A*, path caching, and pitfalls.
Provides multi-turn emotional companion chat through a third-party Agent API and supports continuing sessions via session_id.
An evaluation-driven workflow for diagnosing, optimizing, validating, and logging Agent Skills based on design patterns and anti-patterns.