The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructs the agent to create and manage a hidden credentials file at ~/.agentix/credentials. This file is used to store high-value plaintext secrets, including Agentix API keys, customer IDs, and team identifiers.
[DATA_EXFILTRATION]: The skill explicitly directs the agent to solicit sensitive third-party credentials from the user, specifically Anthropic API keys and GitHub personal access tokens. These secrets are then transmitted via network requests to an external service (agentix.cloud) for storage and use by remote workers.
[PROMPT_INJECTION]: The skill implements a pattern for Indirect Prompt Injection by fetching a 'playbook' from a remote API ($AGENTIX_API/teams/$TEAM_ID/playbook). This remote data is used to define the agent's 'operating mode, policies, and custom rules'.
Ingestion points: Remote content is fetched from the Agentix API and loaded into the agent's context during the initialization of every session.
Boundary markers: The skill lacks explicit boundary markers or instructions to ignore embedded commands within the fetched playbook text.
Capability inventory: The agent has the capability to create roles, manage tasks, spawn remote workers, and interact with GitHub repositories if configured.
Sanitization: There is no evidence of sanitization or validation performed on the remote playbook content before it is adopted as the agent's operating policy.
[COMMAND_EXECUTION]: The skill provides examples of shell commands used for environment setup and credential management, including the use of cat with heredocs to write secrets to the filesystem.

agentix-ceo