The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied campaign details, product descriptions, and external 'Narrative canon' data to generate structured influencer briefs. This ingestion of untrusted or external data creates a surface where malicious instructions could potentially influence the agent's output.
Ingestion points: User inputs (campaign, product, platform, deliverables) and local projection files (e.g., memory/projections/narrative.json, memory/projections/claims.json).
Boundary markers: The skill uses a dependency_status field (verified | approved-fallback | blocked) and markers like [needs source] to track the provenance and validity of claims, which provides some mitigation against accidental obedience.
Capability inventory: The skill has the ability to write files to the local file system (memory/influencer/brief-generator/) and execute the registry-events.py script.
Sanitization: There is no explicit description of sanitizing or escaping the input data before it is interpolated into the brief templates provided in references/brief-templates.md.
[COMMAND_EXECUTION]: The skill instructions specify that unresolved wording should be submitted through registry-events.py as an authorized operation: propose event. This involves the execution of a local script to handle data state changes or proposals within the marketing workflow.

brief-generator