The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes the vendor-specific CLI tool egeo for workspace management tasks, including egeo loop run, egeo loop collect, and egeo loop doctor. It also uses python -m egeo.substrate_lint for artifact validation. These commands are executed locally within the user's environment to maintain the eGEO workspace.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes untrusted data from external sources (web page content and search engine results) which could contain adversarial instructions.
Ingestion points: External data enters the agent context via $EGEO_HOME/data/<collector>/*.jsonl files and the contents of target URLs processed during the 'Do the work' phase.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions to isolate the LLM from potential commands embedded in the collected data.
Capability inventory: The agent possesses the capability to write files, modify the workspace substrate (signals and docs), and execute shell commands via the egeo CLI.
Sanitization: There are no explicit requirements for the agent to sanitize, escape, or validate the content of external data before incorporating it into the workspace or using it to determine the next unit of work.

geo-loop