The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill possesses a high-risk capability to read untrusted local files and write executable configuration files.
Ingestion points: The skill uses ProjectDetector(Path.cwd()) in SKILL.md to scan the local file system.
Boundary markers: Absent; there are no instructions to the agent to ignore instructions or malicious patterns found within the project files.
Capability inventory: The skill uses Write and Bash tools to modify files in .github/workflows/, which are high-privilege files executed by GitHub Actions runners.
Sanitization: Absent; the skill uses engine.render_file to generate output from templates without explicit sanitization of the project-derived metadata.
[Unverifiable Dependencies] (MEDIUM): The Python implementation relies on a non-standard project_detector library and references undefined external commands like /attune:upgrade-project and /pensive:shell-review.
[Dynamic Execution] (MEDIUM): The skill dynamically generates executable YAML workflows at runtime using a template engine (engine.render_file).

workflow-setup