The Agent Skills Directory

[COMMAND_EXECUTION]: The skill guides the agent to execute shell commands using the ngrok CLI and curl for interacting with the ngrok API.
Evidence: Multiple examples in references/cli.md and references/api-resources.md (e.g., ngrok http 8080, curl -X POST https://api.ngrok.com/...).
[INDIRECT_PROMPT_INJECTION]: The agent determines the configuration surface by reading external project files, which creates a potential vector for malicious content in those files to influence agent behavior.
Ingestion points: package.json, pyproject.toml, go.mod, Cargo.toml, *.tf, Chart.yaml, kustomization.yaml, and Kubernetes manifests (identified in SKILL.md).
Boundary markers: Absent; no specific instructions to ignore embedded prompts in these data files.
Capability inventory: Shell command execution via ngrok CLI and network operations via curl to the ngrok.com API (identified across all reference files).
Sanitization: Absent; no mention of sanitizing or validating the content of scanned files.
[PERSISTENCE]: The skill documents the use of ngrok service install, which registers the agent as a persistent background service on the host operating system.
Evidence: ngrok service install --config /etc/ngrok.yml in references/cli.md.
[PRIVILEGE_ESCALATION]: Documented commands for installing and starting system services typically require elevated privileges (root or Administrator) on the target system.
Evidence: references/cli.md explicitly notes that install and start commands for background services "usually need root".

ngrok-surfaces