The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes repository-local shell scripts (./scripts/ai-workflow.sh) and standard system utilities like ps, grep, and crontab to monitor and manage the runtime environment. These commands are used to identify owner leases and verify process integrity.
[INDIRECT_PROMPT_INJECTION]: The skill processes output from local scripts and system commands to determine the state of the autonomous runtime, which serves as a potential surface for indirect injection if these outputs are manipulated by an attacker in the environment.
Ingestion points: Output from ./scripts/ai-workflow.sh, ps -ef, crontab -l, and the graphify CLI.
Boundary markers: None identified; the agent is instructed to interpret script and command output directly to reconcile state.
Capability inventory: Execution of repo-local scripts and system process management via shell.
Sanitization: No explicit sanitization or validation of external script output is described before the agent processes it.
[REMOTE_CODE_EXECUTION]: The skill uses python3 -m pip show graphifyy to verify the version of a dependency. While the package name graphifyy (with double 'y') appears inconsistent with references to graphify elsewhere in the skill, pip show is a diagnostic command that only queries locally installed packages and does not perform remote code execution or installation.

hermes-graphify-gsd-runtime-operator