The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill recommends practices that result in the storage of sensitive Personal Access Tokens (PATs) in plaintext on the local filesystem.
Evidence: Recommends git config --global credential.helper store, which saves credentials to ~/.git-credentials without encryption.
Evidence: Recommends embedding the token directly in the remote URL via git remote set-url origin https://<username>:<token>@github.com/..., which stores the secret in the repository's .git/config file and may leak it into shell history.
[DATA_EXFILTRATION]: The skill and its associated scripts programmatically access sensitive file paths containing authentication secrets.
Evidence: Accesses ~/.git-credentials to extract tokens using grep and sed in SKILL.md and scripts/gh-env.sh.
Evidence: Accesses .env files (specifically ~/.hermes/.env) to read GITHUB_TOKEN values.
Evidence: Performs read operations on SSH private key paths (e.g., ls -la ~/.ssh/id_*.pub).
[COMMAND_EXECUTION]: Uses shell pipelines to parse and extract authentication tokens from files, which can be brittle and involves processing sensitive data through multiple subprocesses.
Evidence: grep "github.com" ~/.git-credentials | head -1 | sed 's|https://[^:]*:\([^@]*\)@.*|\1|' used to programmatically retrieve stored tokens.

github-auth