The Agent Skills Directory

[Data Exposure & Exfiltration] (LOW): The skill intentionally executes commands (npx convex env list --prod, vercel env ls) that output sensitive production credentials into the agent's execution context.
Evidence: Commands in SKILL.md and references/hygiene-checklist.md pipe production secret lists into shell loops for auditing.
Context: This behavior is aligned with the skill's primary purpose of environment variable hygiene and validation.
[Indirect Prompt Injection] (LOW): The skill reads untrusted data from external environment variable stores, which could theoretically contain malicious instructions meant to influence the agent during the 'validation' phase.
Ingestion points: Output of npx convex env list --prod and vercel env ls --environment=production in references/hygiene-checklist.md.
Boundary markers: Absent; the data is processed directly via shell pipelines.
Capability inventory: CLI execution (npx, vercel), shell processing (printf, echo, grep, cut), and cryptographic generation (openssl).
Sanitization: Absent; the skill uses regex for format validation but does not sanitize for executable prompt content.
[Unverifiable Dependencies] (LOW): Relies on standard platform CLI tools (vercel, convex).
Evidence: SKILL.md uses npx convex and vercel command patterns.

env-var-hygiene