The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from ARTIFACT:PLAN using the artifact_read tool. Since this data originates from previous pipeline stages and may contain user-controlled content, it presents an indirect prompt injection surface. The skill performs analysis on this content and interpolates it into a final update via artifact_register without explicit sanitization or boundary enforcement described in the instructions.
[INDIRECT_PROMPT_INJECTION]: Mandatory Evidence Chain:
Ingestion points: The skill reads external content using the artifact_read tool as specified in the Input section of SKILL.md.
Boundary markers: No explicit delimiters or instructions are provided to the agent to treat the ingested artifact content as untrusted data.
Capability inventory: The skill utilizes mcp__plugin_dh_backlog__artifact_register for writing data, mcp__plugin_dh_backlog__profile_list for tool discovery, and requires read access to the entire codebase.
Sanitization: The instructions lack specific sanitization, escaping, or validation steps for the {full_updated_plan_markdown} variable before it is passed to the registration tool.

context-integration