The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external technical plans and GitHub comments to guide implementation, which could contain malicious instructions designed to influence agent behavior during file-writing operations.
Ingestion points: Reads markdown plan files from user-specified paths and fetches GitHub comment threads via desplega:tackle-gh-comments (file: SKILL.md).
Boundary markers: The skill lacks explicit instructions or delimiters to isolate implementation logic from potentially adversarial instructions embedded in the plans or comments.
Capability inventory: The skill utilizes file system write capabilities, spawns sub-agents for phase execution, and executes git commands (file: SKILL.md).
Sanitization: There is no evidence of validation or sanitization of input data before it influences implementation steps.
[COMMAND_EXECUTION]: The skill performs shell command execution and triggers local scripts based on tool usage hooks.
Evidence: Executes git branch --show-current to identify the active development branch (file: SKILL.md).
Evidence: Configures hooks to execute ${CLAUDE_PLUGIN_ROOT}/hooks/plan_checkbox_reminder.py and ${CLAUDE_PLUGIN_ROOT}/hooks/plan_checkbox_stop.py during the session lifecycle (file: SKILL.md).

implementing