The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the use of sudo to start and manage the Docker daemon as described in references/DOCKER.md. It also utilizes shell commands to execute background services via setsid and nohup in references/PREVIEWS.md.
[DATA_EXFILTRATION]: The skill instructs the agent to upload image data to Imgur, a third-party external service, in references/GITHUB.md. It also performs network operations using curl to interact with Slack and Linear APIs, transmitting data authenticated via environment variables (SLACK_BOT_TOKEN, LINEAR_ACCESS_TOKEN).
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its processing of untrusted data from multiple external sources.
Ingestion points: The agent reads data from Slack threads (conversations.replies), Linear issues (issueSearch), and GitHub pull requests or issues (gh pr view, gh issue view).
Boundary markers: There are no specified boundary markers or instructions to delimit or ignore instructions embedded within the ingested data.
Capability inventory: The agent has the ability to execute shell commands (gh, docker, replicas), write to the filesystem (for service logging), and perform arbitrary network requests via curl.
Sanitization: The skill does not mention any sanitization, filtering, or validation steps for content retrieved from external integrations before it is processed by the agent.

replicas-agent