The Agent Skills Directory

[COMMAND_EXECUTION]: The skill implements a comprehensive hooks system that executes local shell and Python scripts (e.g., .claude-organic/hooks/pre-tool-use.py, .claude-organic/engine/flow/initialization.py) to manage workflow states and automate tasks.
[PROMPT_INJECTION]: The report generation functionality in reference/report-output.md represents an indirect prompt injection surface as it ingests untrusted data from Git commit messages and GitHub PRs.
Ingestion points: Data enters via git log, git diff, and gh pr view commands.
Boundary markers: None identified; untrusted commit messages are processed directly into summary reports.
Capability inventory: The skill has broad capabilities, including the ability to write files and execute shell commands through the Bash tool.
Sanitization: No explicit sanitization of commit messages or PR content is documented before processing.
[SAFE]: The skill proactively includes security guards (dangerous_command_guard.py) to detect and block high-risk operations such as rm -rf /, git reset --hard, and chmod 777, enhancing the safety of the agent's environment.

workflow-system