The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Python scripts using shell commands constructed from session variables.
Evidence: The workflow in SKILL.md triggers commands such as uv run python3 $CLAUDE_PROJECT_DIR/tools/brain.py record <target> confirmed "<description>" "<details>".
Security Concern: The variables <target>, <description>, and <details> are sourced from the agent's current session. If this context contains shell metacharacters (e.g., semicolons, backticks, or pipes), it could allow for arbitrary command execution on the host system.
[DATA_EXFILTRATION]: The skill is designed to transmit technical patterns and finding data to an external 'global brain'.
Evidence: uv run python3 $CLAUDE_PROJECT_DIR/tools/global_brain.py learn technique "<pattern>".
Security Concern: This represents a flow of potentially sensitive engagement data to an external repository. While intended for 'cross-target learning,' users should be aware that their activity data is being centralized externally.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the session context which may originate from external targets during auditing.
Ingestion points: SKILL.md instructions specify reading the session context for target, endpoint, and vulnerability details.
Boundary markers: No delimiters or safety instructions are used to isolate untrusted session data from the shell command structure.
Capability inventory: The skill utilizes shell execution capabilities via the uv run command across several scripts.
Sanitization: There is no evidence of sanitization, escaping, or validation of the session-derived strings before they are interpolated into shell commands.

remember