The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill uses npx -y pocketto-pi to download and execute code from the NPM registry during execution. Using the -y flag bypasses confirmation, allowing remote code to run automatically.
[EXTERNAL_DOWNLOADS]: The skill fetches the pocketto-pi package from the public NPM registry. The package is not version-locked, which could lead to the execution of malicious updates or untrusted versions.
[COMMAND_EXECUTION]: The skill executes shell commands via npx that incorporate a user-provided file path (<path-to-execution-plan.md>), which could be leveraged for command injection if the path is not properly handled by the underlying tool.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill reads and processes external execution plans (.md files) and the JSON output from the pocketto-pi CLI.
Boundary markers: None identified. The instructions tell the agent to use the CLI's JSON result as the "source of truth" without explicit validation or delimiters.
Capability inventory: The skill has capabilities for shell command execution (npx), file generation via the CLI, and handoffs to other development skills.
Sanitization: There is no evidence of sanitization or escaping of the content ingested from the execution plan or the CLI output before it is processed by the agent.

pocket-structuring