The Agent Skills Directory

[NO_CODE]: The skill consists entirely of Markdown instructions (SKILL.md), metadata (UPSTREAM.json, PROVENANCE.md), and configuration files (agents/openai.yaml). There are no scripts, binaries, or active code components to execute.
[COMMAND_EXECUTION]: No shell commands or system execution patterns were detected. The disable-model-invocation: true flag in the frontmatter further restricts the agent's ability to call external tools.
[EXTERNAL_DOWNLOADS]: While the skill metadata references an external source (Matt Pocock's GitHub repository) for provenance and audit tracking, the skill itself does not perform any runtime downloads or remote code execution.
[DATA_EXFILTRATION]: There are no network operations or credentials present. The skill operates locally within the provided workspace context (e.g., workflows/*.md and NOTES.md).
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process user input about workflows and store them in local files. While this represents a surface for data ingestion, the skill lacks the capabilities (like code execution or network access) to weaponize such an injection. Severity is safe.

loop-me