The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection by processing untrusted data from local .analysis.md files to drive automated sub-agent workflows.
Ingestion points: In SKILL.md (Step 1.5), the agent extracts Acceptance Criteria (AC) and MECE analysis results from a file following the naming convention {Plan File Path}.analysis.md.
Boundary markers: Absent. The skill lacks robust delimiters or explicit instructions to treat the extracted file content as data rather than instructions, which could lead to sub-agent manipulation.
Capability inventory: The skill delegates tasks to sub-agents with significant capabilities. Specifically, manual-qa-planner.md uses Chrome DevTools MCP tools (navigate_page, fill, click) for browser automation, and auto-qa-planner.md uses Read, Glob, and Grep for filesystem analysis. branch-planner.md also uses Bash restricted to git operations.
Sanitization: Absent. There is no evidence of validation or sanitization of the extracted content before it is interpolated into the prompts for the general-purpose sub-agents.

finalize-plan