The Agent Skills Directory

[SAFE]: The skill follows professional standards for complex AI agent coordination. It includes structured workflows for sub-agent management and explicit normalization steps that act as data validation.
[DATA_EXPOSURE_&_EXFILTRATION]: No evidence of sensitive data access or unauthorized external communication. The skill uses internal routing commands (such as $record and /oh-plan) which are part of the platform's intended design for artifact recording and project planning.
[INDIRECT_PROMPT_INJECTION]: The skill processes session data and sub-agent outputs, creating an attack surface for indirect injection. However, it implements a 'Normalization' step (Step 4) that explicitly filters, flags, and rejects invalid or non-conforming data nodes, which serves as a sanitization mechanism.
Ingestion points: The skill reads the current session's aim and problem space, as well as text blocks returned by child sub-agents during independent framing passes.
Boundary markers: Output is constrained by specific markdown schemas and header structures defined in the workflow.
Capability inventory: The skill can route artifacts to internal storage ($record) and recommend the creation of project tasks/issues via /oh-plan.
Sanitization: The normalization process validates input against S&T (Strategy & Tactics) invariants, rejecting unstructured or suspicious content.
[REMOTE_CODE_EXECUTION]: No remote code execution patterns, package installations, or unauthorized script executions were found.

problem-weave