The Agent Skills Directory

[DATA_EXFILTRATION]: The skill performs sensitive file access to retrieve authentication tokens. It contains logic to read from ~/.git-credentials and $HOME/.hermes/.env to identify and extract existing GitHub Personal Access Tokens.
[DATA_EXFILTRATION]: The instructions recommend storing credentials in plaintext on the filesystem. Method 1 suggests using git config --global credential.helper store, which saves authentication tokens without encryption in the ~/.git-credentials file.
[COMMAND_EXECUTION]: The skill uses shell command substitution and pipe chains to programmatically extract secrets. It employs grep and sed to parse sensitive configuration files and export discovered tokens to the GITHUB_TOKEN environment variable at runtime.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through repository metadata.
Ingestion points: The script scripts/gh-env.sh ingests repository remote URLs directly from the output of the git remote command.
Boundary markers: No delimiters or warnings are used to prevent the agent from obeying instructions that might be embedded in a malicious repository URL or name.
Capability inventory: The skill allows for network operations via curl and gh, and system configuration via git and ssh-keygen.
Sanitization: Input from the git configuration is processed using basic regex in sed and cut without verifying the integrity of the data source or the contents of the parsed strings.

github-auth