The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill uses npx -y pocketto-pi to download and execute code from a remote registry. This execution is triggered automatically if specific headers (like ## Pocket Education) are detected in project documentation files (CLAUDE.md or AGENTS.md).
[EXTERNAL_DOWNLOADS]: The command npx -y pocketto-pi performs an external download of an executable package from the npm registry. The package pocketto-pi is not associated with a trusted organization or vendor, creating a supply chain risk.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources, including project memory files and subagent responses, without sufficient sanitization or boundary markers.
Ingestion points: Reads content from CLAUDE.md and AGENTS.md to check for specific headers, and processes output from general-purpose subagents.
Boundary markers: The prompt templates for subagent dispatch do not include strong delimiters or instructions to ignore potential injection attacks within the plan content.
Capability inventory: The skill has the ability to execute shell commands via npx, read project files, and spawn subagents.
Sanitization: There is no evidence of input validation or escaping for the data read from external files or external agent outputs.

hotfix