The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external and potentially untrusted sources, including image parsing (OCR) and connectors for call recordings, CRM data, and market signals.
Ingestion points: Step 5 (image parsing) and the Connectors section.
Boundary markers: The skill labels pulled facts as drafts and tags them by category, tool, and date to distinguish them from verified data.
Capability inventory: The skill has the ability to read from /foundation/brain.md and write session logs to /context/skill-sessions.md.
Sanitization: A human-in-the-loop requirement is enforced, requiring user confirmation of session setups and discovery findings before generation begins.
[SAFE]: The skill adheres to a rigorous, multi-phase discovery and stress-test methodology based on the Dunford 'Obviously Awesome' framework and includes a mandatory 7-point verification gate to ensure output quality.
[SAFE]: The skill accesses project-specific context files (e.g., /foundation/brain.md) and maintains session logs in the workspace, which are standard operations for context-dependent agent workflows.

positioning-messaging