The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests data from external sources to generate its output.
Ingestion points: Reads plugin metadata and documentation files such as plugin.json, CLAUDE.md, and SKILL.md (as described in the AI Draft Procedure section).
Boundary markers: The skill lacks explicit boundary markers or instructions to the model to ignore potential malicious instructions embedded within the ingested data.
Capability inventory: The skill performs file system reads and writes, and it invokes the mcp__plugin_dh_backlog__backlog_add tool to create external backlog items.
Sanitization: There is no evidence of sanitization or validation of the content read from the plugin files before it is used in prompt construction or output generation.

mission-statement