What the guides are
The short answer
64 guides, 2,048 starters, one structure. A guide groups the starters for one kind of task. A starter is the smallest useful unit: a single job, specified completely enough that an agent can run it without inventing the spec, and a human can review the result against a checklist instead of a vibe.
The index is zambo.dev/workflows/tasks/. Three starters show the range: extract-requirements in the website-brief guide, check-contradictions in the bug-report guide, and handover-current-state in the multi-ai-handover-plan guide.
How a starter is structured
- Inputs. What the task needs before the first call, named up front.
- A copyable prompt. Written for that exact task, ready to paste or load.
- An output checklist. The shape the finished result has to arrive in, item by item.
- Review checks. The questions a reviewer runs before accepting the result.
- Stop conditions. The points where the agent stops and reports instead of pushing on.
- A JSON recipe. The whole starter as structured data, so a tool or another agent can load the job directly.
Each starter carries its own direct anchor link, so the exact job can be shared, saved, or handed to another AI without pointing at a whole page and hoping the reader finds the right section.
Saved, resumed, and shared: the workflow journey
The 25 pre-existing workflows remain at zambo.dev/workflows/, and the index gained four things in the same release.
- A workflow finder. Find by problem or outcome, for when you know the result you need and not the name of the workflow that gets it.
- Browser-saved shortcuts. Saved in localStorage in your own browser. Nothing is sent to Zambo.
- A resume prompt. Pick up where you left off: a job resumes from its saved state in a fresh session or a different AI. Picking up in another AI is always free.
- Public template links. A workflow can be shared as a link, so the next person starts from the same spec.
The document refusal receipt
Document work states its limits on the record. The doc_intel tool analyzes at most 24,000 UTF-16 code units per call and reports exact coverage, the source_sha256, and next_offset on every call.
A clean analysis is public: receipt ac016d46, 992 of 992 characters covered, hash matched. The refusal is public too: if the source changes between parts, the tool refuses with code source_changed and issues a receipt for the refusal, receipt bc83e018, and that refusal consumed no quota.
Run the guides on Zambo
One MCP endpoint runs all of it: zambo.dev/mcp, 153 tools behind one URL, 37 on the default surface, every call returning a verifiable AER-1 receipt. The free tier is 20 calls plus 1 wrap-up free every day, and pickup is always free. The Day Pass is $0.99 for 24 hours, the Zambo Pass is $29 for 30 days ($19 for $ZAMBO holders), and Lifetime is $250, one payment. There is no annual tier.
Questions, answered
What are the Zambo task guides?
64 guides at zambo.dev/workflows/tasks/ holding 2,048 task-specific starter recipes. Each starter is a ready job: inputs, a copyable prompt, an output checklist, review checks, stop conditions, and a JSON recipe, each with its own direct anchor link.
How is a starter different from a prompt template?
A template gives you wording with blanks. A starter gives you the whole job: what goes in, what done looks like as a checklist, how to review it, and where the agent must stop and report. The JSON recipe makes the same job loadable by a tool, not just readable by a person.
What does a Zambo receipt prove?
It proves the saved result was not changed since it was saved. It does not prove the answer was right, and this page will not claim otherwise.