The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's UPSTREAM.md file references source code and documentation from the cursor/plugins GitHub repository to track adaptations and provenance. These references are used for organizational and maintenance purposes.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for reading and explaining arbitrary source code, which constitutes an ingestion surface for indirect prompt injection.
Ingestion points: SKILL.md requires the agent to "Read the relevant source" as the primary input for its teaching task.
Boundary markers: The instructions do not specify the use of markers to delimit untrusted code or provide warnings to ignore instructions embedded within code comments or metadata.
Capability inventory: The skill utilizes tool calls such as /pstack-how, /pstack-why, and /pstack-swarm. Additionally, UPSTREAM.md notes that the agent should "prepare learning-driven source edits for review," indicating file modification capabilities.
Sanitization: No explicit sanitization, filtering, or validation of the ingested source code is described in the workflow.

pstack-teach