The Agent Skills Directory

[COMMAND_EXECUTION]: The skill constructs shell commands by interpolating variables directly into shell strings, such as when creating or superseding records with user-provided titles and basis text. While standard for this toolset, the instructions do not specify sanitization or escaping logic for these inputs, which could lead to command injection if the execution environment does not handle them safely.
[INDIRECT_PROMPT_INJECTION]: The skill processes project-related data from a local store that could be a vector for indirect prompt injection. Ingestion points: In SKILL.md (Step 1), the agent is instructed to read active briefs, ideas, and effort context records. Boundary markers: Procedural steps require explicit human approval or delegation (Steps 4 and 6) before recording artifacts, providing a manual check on the agent's output. Capability inventory: The agent has the capability to modify the local record store and perform automated consistency analysis using orchestra tools. Sanitization: No explicit sanitization or filtering of the ingested store content is mentioned before it is processed or presented.

spec