The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill accesses and reads .env.local to extract secret keys and values, which are categorized as sensitive file paths.
[COMMAND_EXECUTION]: The execution logic uses source .env.local to load variables into the shell environment. This evaluates the file's content as executable shell code, posing a risk if the file contains malicious command substitutions or malformed definitions.
[COMMAND_EXECUTION]: Sensitive values are passed as command-line arguments to the vercel env add command (e.g., --value "$val"). Secrets passed via CLI arguments are potentially visible in system process lists (ps) or shell history logs.
[EXTERNAL_DOWNLOADS]: The skill invokes pnpm exec vercel, which facilitates interaction with Vercel's well-known service infrastructure.
[PROMPT_INJECTION]: The skill reads and processes local project files (env.ts, keys.ts) to reconcile environment schemas, creating a surface for indirect prompt injection.
Ingestion points: grep operations on .env.local, env.ts, and keys.ts in the Diff section.
Boundary markers: None present for the ingested file data.
Capability inventory: Shell command execution via pnpm exec vercel and local file read/write access.
Sanitization: The skill applies strict regex filtering ([A-Z_][A-Z0-9_]*) to extracted variable names before processing.

vercel-env-sync