The Agent Skills Directory

[SAFE]: The skill operates as a specialized workflow architect that facilitates a dialogue to extract requirements from the user. Its behavior is confined to the creation and maintenance of documentation within the workspace.
[DATA_EXPOSURE_AND_EXFILTRATION]: The skill interacts with local files (workflows/*.md and NOTES.md) but lacks any network tools or exfiltration patterns. No hardcoded credentials or sensitive system path accesses were found.
[REMOTE_CODE_EXECUTION]: The skill does not perform package installations, remote script downloads, or shell command execution. The disable-model-invocation setting in the frontmatter further restricts the skill's execution scope.
[PROMPT_INJECTION]: No instructional overrides, DAN patterns, or safety bypass attempts were detected. The "relentless grilling" persona is a functional requirement for the workflow design process and does not target agent behavior modification.
[INDIRECT_PROMPT_INJECTION]: The skill reads and writes workspace data, representing a standard ingestion surface for project-specific information.
Ingestion points: workflows/*.md and NOTES.md are used as the primary data sources for workflow specifications.
Boundary markers: The skill does not explicitly use markers to separate data from instructions within these files.
Capability inventory: The skill is limited to markdown file manipulation (create, edit, delete). It possesses no capabilities for command execution, network requests, or privilege escalation.
Sanitization: No explicit sanitization or validation of the markdown content was identified.

loop-me