The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash tool to execute a Python script via the shell. This execution pattern interpolates user-controlled variables ($SUMMARY_FILE, $PROJECT_DIR_ARG) directly into the script body, creating a potential vector for command injection if input sanitization is bypassed.
[DYNAMIC_EXECUTION]: The skill performs dynamic loading of code by modifying the Python module search path (sys.path.insert(0, '$HOME/.claude/instincts/lib')) to import a local library (gsd_bridge). Loading executable modules from computed paths within the user's home directory is a medium-risk pattern that depends on the integrity of the local filesystem.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and parses external files (GSD artifacts) to extract lessons and patterns, presenting an attack surface for indirect prompt injection.
Ingestion points: The skill reads file content based on the user-provided <summary-file-path> using the Read tool.
Boundary markers: The instructions do not define clear delimiters or provide the agent with instructions to ignore potential commands embedded within the text of the GSD files.
Capability inventory: The skill has high-privilege access including Bash and Write tools, specifically targeting the $HOME/.claude/instincts/ directory.
Sanitization: There is no specified logic for sanitizing or validating the extracted text snippets before they are processed by the bridge library or included in the final report.

instinct-bridge