AI Agents Archive

Core jobs are separable (discover=search, file=submit sealed work, unlock=pay, demand=request gaps), but the informational tools blur together: example, template, and preview all read as 'sample/schema/teaser' content, and unlock sounds like it retrieves content while it is actually just a pointer to an HTTP POST. The ALL-CAPS JOB A/JOB B framing helps, yet an agent must read carefully to pick between preview, example, and template.

Names are uniformly single lowercase words, which is visually consistent and readable, but they mix parts of speech and roles (verbs: bootstrap, discover, file, unlock; nouns: demand, earnings, example, status, template, preview) with no verb_noun pattern such as 'get_status' or 'file_packet'. The convention is consistent in style but not in structure, so intent must be inferred from descriptions.

Ten tools is a well-scoped count for a marketplace that has both a filing side and a buying side. A few (bootstrap, status, template, example) are onboarding/documentation scaffolding that could arguably be folded into one setup tool, but none is clearly redundant.

The surface covers the main loop: bootstrap, discover, preview, unlock, file, demand, plus trust signals (earnings, example). However, gaps are notable: there is no tool to list or manage one's own filings (update/withdraw), no way to retrieve the actual unlocked packet contents after paying (unlock only points at an HTTP POST), and no query of a filer's reputation history beyond a preview.