The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill promotes the storage of authentication secrets in plaintext.
It recommends using the git-credential-store helper (git config --global credential.helper store), which saves GitHub Personal Access Tokens unencrypted in ~/.git-credentials.
It suggests embedding tokens directly in git remote URLs, which results in the secrets being stored in plaintext within the repository's .git/config file.
[DATA_EXFILTRATION]: The skill contains logic to read sensitive local credential files.
The scripts/gh-env.sh script and sections of SKILL.md use grep and sed to parse ~/.git-credentials and extract stored tokens for use in other commands.
[COMMAND_EXECUTION]: The skill provides instructions to create unencrypted private keys.
The ssh-keygen command provided uses the -N "" flag, which generates a private key without a passphrase protection.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface through the ingestion of untrusted repository and credential data.
Ingestion points: The scripts/gh-env.sh script reads data from the environment via git remote get-url and the ~/.git-credentials file.
Boundary markers: There are no boundary markers or instructions to ignore embedded commands when processing these external data sources.
Capability inventory: The skill utilizes curl, git, and ssh-keygen (as seen in SKILL.md and scripts/gh-env.sh) to perform network and file system operations.
Sanitization: The script relies on basic string manipulation (sed, cut) and lacks robust sanitization or validation of the ingested data before it is used in subsequent operations.

github-auth