The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data including interviews, feedback, and support tickets, creating a surface for indirect prompt injection.
Ingestion points: The skill gathers inputs from external sources described in the 'Procedures' and 'Input Types' sections of SKILL.md (interviews, tickets, research, analytics).
Boundary markers: No specific delimiters or boundary markers are instructed for the agent to distinguish between its instructions and the processed content.
Capability inventory: The skill generates local markdown files (DISCOVERY_LOG.md, EXPERIMENTS.md) and passes instructions to downstream agents (e.g., requirement-architect). It does not perform network operations, subprocess execution, or arbitrary code execution.
Sanitization: There are no instructions for sanitizing or escaping the content retrieved from external sources before it is interpolated into the agent's context.
[NO_CODE]: The skill consists entirely of markdown instructions and templates. No executable scripts (Python, JavaScript, shell, etc.) are included in the skill definition.
[SAFE]: The skill includes a mandatory 'Human Gate 0' (Discovery Approval) which requires explicit user validation of findings and candidate requirements before proceeding, significantly reducing the risk of automated exploitation of the agent's logic.
[SAFE]: The 'Traceability Rule' and 'Halt Conditions' enforce a strict lineage requirement, ensuring that every proposed requirement must be tied back to a discovery artifact or a specific stakeholder directive.

discovery-analyst