The Agent Skills Directory

[SAFE]: The skill implements a standard software development recovery workflow focusing on assessment, evidence preservation, and rollback.
[DATA_EXFILTRATION]: The skill captures environment variables using env | grep in SKILL.md, but restricts the output to specific prefixes (NODE, NPM, PNPM, PATH), which is a safe practice for debugging without exposing sensitive credentials.
[EXTERNAL_DOWNLOADS]: The skill uses pnpm install --frozen-lockfile and pnpm dlx madge. These are well-known package management operations used within their intended development context.
[COMMAND_EXECUTION]: The skill utilizes the GitHub CLI (gh issue comment) to document recovery efforts. This is a legitimate use of developer tooling for incident reporting.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data such as error logs and stack traces, which represents a potential attack surface.
Ingestion points: Command outputs from pnpm test, git status, and env are captured in SKILL.md for assessment.
Boundary markers: Captured content is wrapped in markdown code blocks to separate it from instructions.
Capability inventory: The skill has access to Bash execution, Git operations, and GitHub CLI tools.
Sanitization: None explicitly defined; logs are interpolated into markdown templates, though typically within code blocks.

error-recovery