The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various system commands including git diff, git log, and open to reconstruct session history and open generated visual reports.
[DYNAMIC_EXECUTION]: The skill automatically generates and writes Python and Bash scripts to ~/.claude/hooks/. This allows for the creation of executable content based on session summaries, which can be risky if the input data is influenced by malicious content.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from session context, git logs, and diffs to determine what artifacts to create.
Ingestion points: git diff, git log, and conversation context.
Boundary markers: None specified to distinguish between developer intent and data content within the logs.
Capability inventory: Writing .sh and .py hooks, modifying settings.json, and executing system commands.
Sanitization: No explicit sanitization of session data is performed before it is used to generate configuration or scripts.
[DATA_EXPOSURE]: The skill reads and modifies sensitive agent configuration and memory files located in ~/.claude/, including settings.json, agents/, and skills/ directories.

done