The Agent Skills Directory

[PROMPT_INJECTION]: The skill uses instructional constraints like and strict procedural checklists to enforce a specific workflow. These directives are functional and do not attempt to bypass core safety guardrails or manipulate the agent's underlying behavior instructions.\n- [DATA_EXFILTRATION]: The skill accesses project-specific files such as docs/constitution.md and docs/ledger.jsonl to maintain context. These actions are limited to local file reading for intended functionality and do not involve sensitive system credentials or network operations to external domains.\n- [COMMAND_EXECUTION]: There are no patterns of arbitrary shell command execution, subprocess spawning, or system configuration changes. The skill's primary actions are conversational or file-based (writing design documents).\n- [REMOTE_CODE_EXECUTION]: The skill does not download or execute remote code. While it references other skills (e.g., evo-writing-plans), these are internal transitions within the vendor's ecosystem and do not involve untrusted third-party dependencies or remote script execution.

evo-brainstorming