The Agent Skills Directory

[SAFE]: No security issues, obfuscation, or malicious patterns were detected. The skill focuses on structured logging of development experience.
[COMMAND_EXECUTION]: The skill utilizes git diff to analyze code changes. This is a standard local operation appropriate for a development-focused agent.
[DATA_EXPOSURE]: The skill reads internal project state files from the .ai_state/ directory and git history. These operations are conducted locally, and no network exfiltration or credential harvesting patterns are present.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from git diff and sprint reports.
Ingestion points: Reads git diff, .ai_state/reviews/sprint-N.md, .ai_state/tasks.md, and .ai_state/design.md.
Boundary markers: The instructions enforce a strict structured output format for the lessons.md file, though specific delimiters for the input data are not explicitly defined in the prompt text.
Capability inventory: Limited to local file read/append and git diff execution; no network access or arbitrary code execution tools are requested or used.
Sanitization: The skill uses a distillation method that focuses on specific categories (Pattern, Pitfall, Constraint), which naturally filters and summarizes the input rather than reflecting it raw.

compound