The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several local bash scripts located in ${HOME}/.claude-octopus/plugin/scripts/. These include octo-state.sh, state-manager.sh, check-providers.sh, and orchestrate.sh. These scripts are part of the 'Claude Octopus' plugin and are used for state management and workflow orchestration.
[EXTERNAL_DOWNLOADS]: The skill references external providers (Codex, Antigravity) but interacts with them through local CLI tools. No direct remote code execution or untrusted downloads were detected.
[CREDENTIALS_UNSAFE]: The skill mentions the use of OPENAI_API_KEY and AGY_AUTH_TOKEN in its cost awareness section, but it does not hardcode them or instruct the agent to exfiltrate them. It assumes they are already configured in the environment for the respective CLI tools.
[DYNAMIC_EXECUTION]: The skill uses python3 to run a local script search.py for design intelligence if the vendor plugin ui-ux-pro-max-skill is present. This is a local execution of a component from the same vendor ecosystem.
[PRIVILEGE_ESCALATION]: All commands are executed within the user's home directory context. No use of sudo or other privilege escalation techniques was found.
[INDIRECT_PROMPT_INJECTION]: The skill processes user prompts and project files (like package.json). It implements a 'Quality Supplement' enrichment layer to inject specific constraints into the prompt sent to external providers, which acts as a safety and quality boundary.

flow-develop