The Agent Skills Directory

[NO_CODE]: The skill is composed exclusively of Markdown documentation (SKILL.md, references/experiments.md, references/form.md), evaluation definitions (evals/evals.json), and licensing/metadata files. No executable scripts, binaries, or active code components are included in the distribution.
[SAFE]: No sensitive file access or credential exposure patterns were detected. The skill instructions mention reading project-specific context files such as .agents/product-marketing.md or .claude/product-marketing.md, which is a standard procedure for retrieving local project documentation and does not target system-level secrets (e.g., SSH keys, environment variables).
[SAFE]: No obfuscation, persistence mechanisms, or dynamic execution patterns were found. The skill does not utilize the !command syntax for dynamic context injection.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to analyze external URLs provided by the user to perform conversion rate optimization audits. This creates a surface where a target page could contain hidden instructions intended to influence the agent. However, the skill lacks any dangerous capabilities—such as file writing, shell execution, or network exfiltration—that could be leveraged by such an injection. The agent is strictly limited to providing text-based analytical feedback.
Ingestion points: User-provided URLs for page analysis (SKILL.md).
Boundary markers: None; the skill does not explicitly instruct the agent to use delimiters or isolate the content of the analyzed URL.
Capability inventory: None; no tools for file system modification, command execution, or external network requests are enabled or requested.
Sanitization: None; the content is ingested as raw text for analysis.

marketing-cro