The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to download a pre-built binary from a personal GitHub Pages site and a personal GitHub repository.
Evidence: Downloads from https://randyfajar.github.io/inbox-watcher-cli/ and https://github.com/randyfajar/inbox-watcher-cli/releases/latest/download/chronosforge-linux.
[COMMAND_EXECUTION]: The skill provides instructions and examples for executing the downloaded binary, including the use of administrative privileges and shell subprocesses.
Evidence: sudo mv chronosforge /usr/local/bin/ in the GitHub Actions example. Integration examples use subprocess.run (Python) and exec (Node.js) to call the tool.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests and processes untrusted data from external email messages.
Ingestion points: Processes incoming emails and message bodies via the chronosforge inbox watch and chronosforge extract commands.
Boundary markers: None explicitly defined in the prompt instructions to isolate external email content from the agent's logic.
Capability inventory: The skill allows for local file access (config files), network operations (API/Webhooks), and command execution via integration scripts.
Sanitization: While it uses regex patterns for code extraction, the chronosforge inbox message <id> --raw command allows for the processing of unsanitized external content.

chronosforge-temporal-email-cli