The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the execution of multiple shell scripts located in a specific local directory (~/.claude-octopus/plugin/scripts/) to check provider status, manage state, and orchestrate external AI models.
[DYNAMIC_EXECUTION]: The skill instructions direct the agent to construct and execute shell commands using variables populated from user responses. Specifically, in Step 3, variables such as user_flow and approach are interpolated into shell commands, and in Step 4, the user's clarification request is placed directly into a bash command string (orchestrate.sh define "<user's clarification request>"). A user providing input containing shell metacharacters like backticks or semicolons could trigger arbitrary command execution within the agent's environment.
[INDIRECT_PROMPT_INJECTION]: The workflow involves reading and synthesizing output from external AI providers (Codex and Antigravity), creating a vulnerability surface for indirect prompt injection.
Ingestion points: The skill reads discovery findings (discover_context), prior decisions, and the multi-AI synthesis results (grasp-synthesis-*.md) from the file system (SKILL.md Steps 2 and 5).
Boundary markers: Absent. The skill does not employ delimiters or instructions to treat data from external models as untrusted or to ignore embedded instructions.
Capability inventory: The agent is granted capabilities for shell command execution and file system read/write access.
Sanitization: Absent. There are no mechanisms defined to validate or escape the content ingested from external providers or user input before it is used in the synthesis process or subsequent commands.

flow-define