The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes content from external source files and documentation to determine its actions.
Ingestion points: The agent reads .doc-sentinel-drift.json in Phase 1 and arbitrary source and documentation files in Phase 3a to identify changes.
Boundary markers: The instructions do not provide delimiters or specific guidance to the agent to treat the content of these files as passive data, potentially allowing embedded instructions in the code or docs to influence agent behavior.
Capability inventory: The skill has access to the Bash tool (for git and file operations), Write, Edit, and the Agent tool for complex resolution tasks.
Sanitization: No sanitization, validation, or escaping of the content read from files is performed before it is analyzed by the agent.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform local file system operations and repository management.
Evidence: The skill executes cat to read state, git add and git commit to persist documentation fixes, and rm -f to clean up the drift tracking file. These operations are aligned with the skill's stated purpose.

resolve