The Agent Skills Directory

[SAFE]: The skill implements a robust state machine via a progress.md file to manage multi-phase workflows. This architecture ensures that operations are resumable and prevents the unintentional loss of work or silent overwriting of finalized plans.
[DATA_EXFILTRATION]: The skill's instructions involve reading data from standard marketing and business APIs such as Stripe, Google Analytics, and Ahrefs. These operations are scoped to the intended purpose of market research and financial auditing, with data persisted in a local research.md file within the client's project folder.
[COMMAND_EXECUTION]: The operational stack described in the references includes the use of various CLI tools and MCP integrations (e.g., agent-browser, defuddle, and gh). These are legitimate developer and marketing tools used to automate data gathering and content management, and their use is consistent with the skill's stated goals.
[SAFE]: The skill demonstrates safe handling of external data by including a mandatory 'REVIEW' phase. In this phase, the agent must present each drafted section to the user for interactive approval or adjustment before it is finalized, providing a manual checkpoint against potential indirect prompt injections from client materials.

marketing-plan