The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides explicit instructions for executing files and binaries within the AFS environment using the arc afs exec command. It details how to pass complex JSON arguments to these executables, granting the agent the ability to run arbitrary code located at specific virtual paths.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest data from the AFS environment through commands like arc afs read, arc afs search, and arc afs explain. This creates a vulnerability surface where malicious instructions embedded in AFS files could influence the agent's behavior.
Ingestion points: arc afs read <path>, arc afs search <path> <query>, and arc afs explain [topic|path]. (Found in SKILL.md)
Boundary markers: No specific boundary markers or instructions to ignore embedded commands are provided.
Capability inventory: The agent has the capability to write files (arc afs write), execute binaries (arc afs exec), and manage UI sessions (arc afs exec /dev/ui/web/.actions/...). (Found in SKILL.md)
Sanitization: No sanitization or validation of the ingested content is mentioned.
[PRIVILEGE_ESCALATION]: The skill includes instructions for using arc attach to connect to local daemon processes and arc afs mount to modify configuration files (.afs-config/config.toml). These commands allow the agent to manipulate the execution environment and how virtual paths are mapped to physical storage.

afs