The Agent Skills Directory

[SAFE]: The skill operates entirely as a text-based creative workflow. It provides natural language instructions for the AI to act as a brand strategist, with no executable code, tool invocations, or network activity.
[NO_CODE]: The provided files consist of markdown instructions (SKILL.md) and evaluation test cases (evals.json). There are no scripts (Python, JavaScript), binaries, or shell commands included in the skill package.
[PROMPT_INJECTION]: No evidence was found of instructions attempting to bypass safety filters, extract system prompts, or override agent behavior. The instructions are focused on the domain of brand strategy.
[DATA_EXFILTRATION]: The skill does not contain any patterns for accessing sensitive files (e.g., .ssh, .aws, .env) or hardcoded credentials. There are no network tools or commands to transmit data to external domains.
[PROMPT_INJECTION]: Regarding indirect prompt injection (Category 8): 1. Ingestion points: The skill processes user-supplied questionnaire responses or briefs in 'STEP 3' to generate reports. 2. Boundary markers: The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings for user data. 3. Capability inventory: The skill's capabilities are limited to text generation; it has no access to subprocesses, file-writing tools, or network operations. 4. Sanitization: No validation or sanitization of input data is defined. The risk is assessed as negligible because the skill lacks the capabilities required for an injection to cause harm beyond text generation.

brand-strategy