The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from the local project environment to generate reports and determine build health.
Ingestion points: The skill reads content from project.yaml, production/qa/qa-plan-*.md, sprint plans, and automated test runner outputs (logs and XML results).
Boundary markers: No explicit delimiters or instructions are provided to the agent to treat this ingested content as data rather than instructions, which could allow maliciously crafted project files to influence agent behavior.
Capability inventory: The skill uses the Bash tool to execute various commands and the Write tool to create report files.
Sanitization: There is no evidence of data sanitization or validation of the external inputs before they are interpolated into the agent's context.
[DYNAMIC_CONTEXT_INJECTION]: The skill utilizes load-time shell execution to retrieve project configuration settings.
Evidence: The skill uses the ! syntax to execute yaml-helper.sh at load time: !bash "${CLAUDE_SKILL_DIR}/../../hooks/yaml-helper.sh" resolve_config --keys automation,qa.level,testing.strict.
This represents a powerful mechanism that runs before the AI processes the skill body. In this case, it appears to be a legitimate use of project-specific tooling for configuration management.

smoke-check