The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from multiple untrusted or external sources (local PRD files, Figma design nodes, and git diffs/logs) and uses this content to generate structured test cases. There is a risk that malicious instructions embedded within these requirements or code comments could influence the agent's output or test generation logic.
Ingestion points: Reads ~/.ai-software-agency/projects/<slug>/prd.md, fetches Figma design context via MCP, and executes git diff on external branches.
Boundary markers: The skill lacks explicit boundary markers or instructions to ignore embedded prompts within the ingested PRD or code diff content.
Capability inventory: The skill has the capability to write to the local filesystem (tests.md, audit.log) and execute shell commands (git).
Sanitization: No explicit sanitization or filtering of the ingested requirement text is performed before interpolation into the generation steps.
[COMMAND_EXECUTION]: The skill executes shell commands (git fetch, git log, git diff) using user-provided branch names. While this is necessary for its function as a QA tool, it represents a standard command execution surface.
[SAFE]: The skill implements a 'Governance Gate' (Step 7), requiring explicit user approval before writing any generated content to the filesystem, which serves as a significant manual mitigation for potential injection or malicious generation.

testcase-generate