The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructions do not contain any patterns attempting to override agent safety guidelines, extract system prompts, or bypass ethical constraints. The instructions focus on establishing rules for honest and specific outreach.
[DATA_EXFILTRATION]: No network operations (such as curl or wget) or sensitive file path accesses were detected. The skill instructions mention tracking fields (e.g., contact names, emails), which is standard for the described use case of managing outreach campaigns.
[EXTERNAL_DOWNLOADS]: The skill does not define any external dependencies or remote resource fetches. All content is contained within the local markdown files.
[REMOTE_CODE_EXECUTION]: There are no scripts, binaries, or commands that execute code from remote sources. The skill is purely informational.
[COMMAND_EXECUTION]: No shell commands or subprocess calls are present in any of the skill files. The skill does not utilize the dynamic context injection syntax (!command).
[INDIRECT_PROMPT_INJECTION]: The skill describes an attack surface where untrusted data (prospect sites, articles, and linkable assets) is ingested by the agent to generate drafts. However, it provides a comprehensive 'Email Review Rubric' in references/core/rules.md to ensure the generated content remains honest and compliant, mitigating accidental obedience to malicious instructions embedded in processed web content.

link-outreach-acquisition