The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands like git log, git branch, and grep to gather information about the project's state. These are read-only operations for context discovery and do not involve high-risk commands or privilege escalation.
[SAFE]: A critical safeguard is implemented in Step 7, which requires explicit user approval ('Czy zatwierdzić te zmiany w dokumentacji?') before the agent is permitted to stage or commit changes to the repository. This prevents unauthorized or autonomous modifications to the project's source or documentation.
[PROMPT_INJECTION]: The skill identifies a potential surface for indirect prompt injection as it ingests and processes existing markdown files and git logs from the repository.
Ingestion points: Reads project documentation (.md files), PROGRESS_TRACKER.md, and CLAUDE.md.
Boundary markers: Not explicitly defined for the read operations, though the skill is designed to condense and remove content rather than interpret instructions within the text.
Capability inventory: Accesses file system (read/write), project metadata (git), and search utilities (grep, glob).
Sanitization: No specific text sanitization is mentioned, but the mandatory user review of the final output acts as a primary control mechanism.

doc-update