The Agent Skills Directory

[DATA_EXFILTRATION]: The skill explicitly targets sensitive files in the user's home directory (~/.claude/CLAUDE.md). Accessing user-level configuration files outside the current project workspace constitutes unauthorized data exposure.
[COMMAND_EXECUTION]: The skill performs broad recursive file system scanning ({repo}/**/AGENTS.md) and probes for various configuration files across the user's system to detect and read their contents.
[PROMPT_INJECTION]: The skill is highly susceptible to indirect prompt injection. It ingests data from external, potentially attacker-controlled sources and injects it into the session context.
Ingestion points: ~/.claude/CLAUDE.md, {repo}/AGENTS.md, {repo}/CLAUDE.md, and any markdown files found within .memory/ or memory-related directories.
Boundary markers: Absent. The skill instructions do not specify any delimiters or warnings to ignore instructions embedded within the retrieved knowledge.
Capability inventory: Extensive file-read capabilities across the project repository and the user's home directory.
Sanitization: Absent. The skill does not perform any validation, filtering, or escaping of the content retrieved from the file system before presenting it to the agent.

knowledge-fetch