The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes acceptance criteria and test case descriptions from repository files to generate reports, creating a surface for embedded instructions to influence agent behavior.
Ingestion points: README.md files (user stories, work items, features) and the test-cases/ directory.
Boundary markers: The skill does not implement delimiters or explicit instructions to distinguish between data and commands when reading these files.
Capability inventory: The skill performs file writes (trace-report.md), executes shell commands via git, and delegates testing tasks to the quality-check tool.
Sanitization: No explicit sanitization or validation of the ingested text content is described in the workflow.
[COMMAND_EXECUTION]: The skill relies on shell commands to generate file fingerprints for its idempotency and caching mechanism.
Evidence: Use of git ls-files, git status, git diff, and git hash-object in references/flow.md to compute FINGERPRINT and SPEC_FINGERPRINT.

trace-validate