The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands to verify technology versions and compatibility.
Evidence: In references/version-verification.md, the agent is directed to use commands such as npm view <pkg> dist-tags --json and pip index versions <pkg> to retrieve real-time data from registries.
[EXTERNAL_DOWNLOADS]: The skill performs network operations to fetch metadata from established public registries.
Evidence: The skill queries registry.npmjs.org and pypi.org as part of its version verification workflow, as documented in references/version-verification.md. These are well-known services used for legitimate developer tooling.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface where user-supplied inputs are used to form shell commands without explicit sanitization instructions.
Ingestion points: The skill ingests technology and package names provided by the user during the selection process (Phase 1 in SKILL.md).
Boundary markers: There are no instructions for using delimiters or boundary markers to prevent the agent from misinterpreting malicious strings embedded in user inputs.
Capability inventory: The agent has the capability to execute shell commands (npm, pip, curl) and perform file system writes (docs/stack/decision.md, AGENTS.md).
Sanitization: No sanitization, escaping, or validation logic is defined in the instructions for the interpolated package names used in the diagnostic commands.

fullstack-stack-guide