The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several shell commands to detect environment versions and analyze project configurations.
Evidence: Running npx tsc --version and node -v in SKILL.md.
Evidence: node -e is used in SKILL.md to parse package.json for dependency detection.
Evidence: scripts/ts_diagnostic.py uses subprocess.run(shell=True) to execute diagnostic commands such as npx tsc --noEmit.
[EXTERNAL_DOWNLOADS]: The skill utilizes npx to invoke tools from the TypeScript ecosystem, which may involve downloading packages at runtime.
Evidence: References to tsc, vitest, tsx, ts-node, ts-migrate, and typesync via npx in SKILL.md and scripts/ts_diagnostic.py.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its processing of untrusted data from local project files.
Ingestion points: Reads package.json, tsconfig.json, and scans the src/ directory as part of its diagnostic routine.
Boundary markers: No explicit delimiters or instructions are used to separate untrusted project data from agent instructions.
Capability inventory: The skill can execute shell commands via Python's subprocess and through npm/npx instructions.
Sanitization: Data read from configuration files is not sanitized before it is used or displayed by the agent.

typescript-expert