SKU 1 · One-time purchase
SKILLSTATE Long-Job Pack
Keep a long agent job resumable: current candidate, unfinished checks, and the next authorized action — in files, not chat memory. Editable Markdown — no service, subscription, installer, or API key. First setup is a suggested working block, not a measured completion claim. No earnings or hours-saved promise.
$49
One payment. Instant download. 14-day refund policy.
Not sure the method fits? Use the free starter card tonight (no email required), or grab it from the magnet page if you want a copy emailed. Buy this pack when the job will cross days or seats.
What you can do after you buy
- Set up a bounded job from the README without a human explainer.
- Refuse an unsupported “done,” keep the failed check, and invalidate stale evidence.
- Recover an interrupted save from artifacts instead of retrying a send or charge.
- Hand the job to a new seat that can name the current candidate and the next safe action.
- Start a second job from the blank master a week later — that is the reuse bar.
Those are procedures in the zip. They are not measured buyer results and not a guarantee the work is correct. Linked evidence has to be.
What is in the zip
README.md
Operating manual: first setup, the loop, resume/recover/land, and what this is not.
OPERATOR_NOTES.md
Week-1 desk card: second job from the blank master, Monday resume, compaction, seat friction.
IMMUTABLE_SPEC.md
Versioned job contract (P): outcome, authority, boundaries, acceptance tests, budget.
MUTABLE_STATE.md
Current-truth record (Σ): one labeled starter JSON object, plus a clearly marked update fragment that is not a complete state.
JOURNAL.md
Append-only decision and recovery trail. Retrieve named entries; do not replay the whole log every turn.
HANDOFF.md
Seat-transfer protocol: current candidate, unproven work, first safe action, receiver acknowledgement.
CHECKLIST_LONG_JOB.md
Preflight, midflight, pause/crash/transfer, landing, and week-1 reuse checks. Ships unchecked.
AGENT_PROMPT.md
Pasteable start/resume and repair instructions. Text you give the agent; not an installer.
ANTI_PATTERNS.md
Symptoms (silent scope drift, fake “90% done”, duplicate sends) and the repair at the next checkpoint.
USE_AND_LIMITS.md
Included rights, practical load notes, and the compatibility evidence boundary.
EXAMPLES/example-agent-desk.md
Authored three-day desk example with a complete state snapshot. Teaching material, not a customer result.
EXAMPLES/example-ops-research.md
Authored research example with incomparable sources and a complete state snapshot. Teaching material.
LICENSE.md
One seat, unlimited jobs, personal or commercial use of the templates. Do not resell the blank pack. Refunds follow /policy.
Who it is for
People running agent jobs that outlast a single sitting — research desks, migrations, long refactors, anything handed between sessions or between agents. If your jobs finish in one prompt, the free card is enough. This pack does not make the work correct; linked evidence has to.
Questions
What format is it?
A zip of thirteen markdown files. Open them in any editor, paste prompts into any agent, commit the copies you fill in to any repo you are authorized to use.
Which agent or model does this need?
None in particular. The pack is editable text. Confirm in your own environment that the agent can actually read and reopen the files. No end-to-end runtime test is claimed for this edition.
How is this different from the free card?
The card is one combined job note plus a short prompt for a bounded job tonight. The pack splits the records into separate files and adds the operating manual, operator notes, start/resume/repair prompts, handoff protocol, checklist, anti-patterns, use/limits, and two authored examples.
Do I get updates?
Your receipt page remints a download of whatever version this storefront is currently shipping. There is no subscription, no second charge, and no promise of future editions.
Path: free card first, this pack if the job will still be open next week. Not a guarantee of earnings, model performance, or time saved. Refunds follow /policy.