The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process team-shared AI skills, rules, and documentation. This architecture creates an attack surface where malicious instructions embedded in a team's shared knowledge base could influence the agent's behavior.
Ingestion points: Data enters the agent context via the teamai skill get and teamai skill list commands, which retrieve external content.
Boundary markers: There are no specified delimiters or 'ignore' warnings for the fetched content described in the documentation.
Capability inventory: The skill possesses the ability to execute Bash commands via the teamai CLI and npx.
Sanitization: The discovery stub does not describe any validation or filtering for the external content it retrieves.
[DYNAMIC_EXECUTION]: The skill utilizes a 'discovery stub' pattern where the actual operational instructions are not contained within the static SKILL.md file. Instead, they are fetched and loaded at runtime using subcommands like teamai skill get core. This dynamic loading mechanism allows the agent's core logic to change based on external tool output.
[EXTERNAL_DOWNLOADS]: The skill directs the user to install the teamai-cli package globally from the official NPM registry using npm i -g teamai-cli@latest. As this targets a well-known service and the package is associated with the skill's author, the download itself is standard functionality.
[COMMAND_EXECUTION]: The skill relies on the execution of shell commands through the Bash tool to perform its primary functions, including skill retrieval, status checking, and environment management.

teamai