The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/agora.py uses subprocess.run to execute various Git commands (such as rev-parse, symbolic-ref, and git-common-dir) to identify repository identity and worktree state. The arguments passed to these commands are generally hardcoded strings or sanitized using a slugging function.
[PROMPT_INJECTION]: The skill facilitates indirect prompt injection by design, as it instructs the agent to read and follow coordination messages that may be authored by other agents or humans.
Ingestion points: scripts/agora.py reads message files from the ~/.agora/ directory via the read, inbox, watch, and tail commands.
Boundary markers: While messages are rendered with metadata headers (timestamp, handle, message type), the body of the message is not enclosed in explicit boundary delimiters or 'ignore embedded instructions' warnings.
Capability inventory: The agent utilizing this skill typically has capabilities to read, write, and modify files within the repository and execute shell commands.
Sanitization: There is no sanitization or escaping of the Markdown content retrieved from the agora before it is presented to the agent's context.
[SAFE]: The skill operates entirely on the local filesystem (defaulting to ~/.agora) and does not perform network operations, download remote code, or require third-party dependencies.

agora