The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill operates on project-level configuration files which could be manipulated by an attacker to include hidden instructions in fields like descriptions or scripts.
Ingestion points: package.json, pnpm-workspace.yaml, and .npmrc files are accessed as described in SKILL.md and references/workspaces.md.
Boundary markers: The documentation does not provide specific instructions to the agent on how to delimit or ignore potentially malicious content within these manifest files.
Capability inventory: The skill facilitates powerful operations including arbitrary shell script execution via pnpm run, binary execution via pnpm exec, and network-based package retrieval as detailed in references/cli.md.
Sanitization: There is no evidence of sanitization or validation of data extracted from project manifests before it is used in commands.
[DYNAMIC_EXECUTION]: The skill documents the usage of .pnpmfile.cjs in references/features.md. This configuration file allows developers to execute arbitrary JavaScript logic during the dependency resolution phase. While a standard feature of the tool, it represents a mechanism for dynamic execution of local script content during the agent's workflow.

pnpm