The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes several management and hook scripts (extract-skill.sh, activator.sh, error-detector.sh) that use shell utilities to process data. Specifically, extract-skill.sh creates new directory structures and files using mkdir, sed, and awk. These scripts include appropriate input validation (e.g., regex checks on skill names) to prevent command injection or path traversal.
[PROMPT_INJECTION]: The skill implements an indirect prompt injection surface (Category 8) by design, as it instructs the agent to review and adhere to patterns and rules logged in the .learnings/ directory.
Ingestion points: Files located within the .learnings/ directory, including LEARNINGS.md, ERRORS.md, and FEATURE_REQUESTS.md.
Boundary markers: Absent; the instructions do not specify the use of delimiters or 'ignore' instructions when reading stored learnings, which could allow malicious instructions in logs to influence agent behavior.
Capability inventory: The skill provides the agent with file-system write access, directory creation capabilities, and shell script execution through the provided hook scripts.
Sanitization: Absent; the content retrieved from the logs is processed as instructions for future tasks without validation or filtering.

self-improvement