The Agent Skills Directory

[SAFE]: The skill instructions define a purely interactive session for gathering and refining workflow requirements. No malicious instructions or hidden behaviors were found in the documentation or configuration files.
[NO_CODE]: The skill package does not include any scripts (Python, Node.js, Shell) or binaries. Its functionality relies entirely on the agent's natural language processing within the provided constraints.
[INDIRECT_PROMPT_INJECTION]: Although the instructions mention reading from workspace files like NOTES.md—which could contain untrusted data—the potential for exploitation is neutralized by the skill's configuration.
Ingestion points: Workspace files NOTES.md and workflows/*.md.
Boundary markers: Absent in the instructions.
Capability inventory: No actionable tool capabilities; disable-model-invocation: true prevents the agent from calling any file-system or network tools.
Sanitization: Not specified, as output is intended for human review.

loop-me