The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands to perform repository management and automation tasks.\n
Uses git checkout, git pull, git add, and git commit to manage phase-specific branches and update status files.\n
Utilizes the GitHub CLI (gh pr create) to programmatically generate pull requests for completed work phases.\n- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing untrusted markdown data to drive agent behavior.\n
Ingestion points: Extracts task descriptions and phase definitions from openspec/changes/<change-id>/tasks.md and proposal.md as described in Step 1.\n
Boundary markers: There are no explicit delimiters or specific instructions to ignore malicious directives embedded within the task descriptions when they are passed to the subagent workflow.\n
Capability inventory: The skill possesses capabilities for filesystem writes, git operations, and triggering the superpowers:subagent-driven-development skill, which likely includes code execution permissions.\n
Sanitization: Content from the markdown files is interpolated directly into task plans for subagents without validation or escaping, allowing potentially malicious task descriptions to influence subagent actions.

openspec-dev