The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided context from NOTES.md, which creates a theoretical attack surface for indirect prompt injection. Malicious instructions embedded in these notes could attempt to influence the agent's behavior during the "grilling" process. The impact is minimal because the skill is confined to the local workspace and cannot invoke external tools or network services.
Ingestion points: The skill reads from NOTES.md to establish user context as defined in SKILL.md.
Boundary markers: Absent; there are no instructions to differentiate user data from system commands.
Capability inventory: The skill is restricted to reading and writing workflows/*.md and NOTES.md files; no subprocess calls or network operations are present.
Sanitization: Absent; no validation or filtering is performed on the ingested content.

loop-me