The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest findings from external sources and store them in a persistent project file, creating a surface for indirect prompt injection.
Ingestion points: Processes output from other agents such as the design-critic and accessibility-reviewer, as well as direct user requests for debt status and summaries.
Boundary markers: Utilizes structured Markdown tables and specific section headers (## Design Debt Register) to delimit data, which provides basic structural separation but does not prevent malicious instructions within the data fields.
Capability inventory: The skill facilitates reading and writing to the local design-state.md file. No high-risk capabilities such as network exfiltration, arbitrary command execution, or system-level configuration changes were detected.
Sanitization: The instructions do not specify validation or sanitization routines for the descriptive content imported from other agents before it is appended to the register.

design-debt-tracker