The Agent Skills Directory

[PROMPT_INJECTION]: The skill utilizes the $ARGUMENTS variable to define the scope of the analysis, which allows user-provided input to directly influence the agent's task definition.
[DATA_EXFILTRATION]: The skill is designed to investigate sensitive configuration files (e.g., package.json, requirements.txt), database connection details, and authentication mechanisms within a target codebase.
[DATA_EXFILTRATION]: The 'Self-Evolution Telemetry' phase instructs the agent to append execution data to ~/.claude/projects/skill-telemetry.md, which involves accessing and modifying a file within a hidden directory in the user's home folder.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its automated analysis of untrusted files. Ingestion points: Entire codebase including source code and configuration files. Boundary markers: None identified; the agent is instructed to investigate the codebase thoroughly without markers. Capability inventory: Filesystem read access and the ability to append to a local telemetry log. Sanitization: No instructions are provided to sanitize or disregard embedded directives within the analyzed files.

legal-aid