Plan: Launch with proof
Written for an AI agent to fetch and follow: it says what to connect, which steps to build with which product, what a person must approve, and how to check the result. Steps that depend on a part that is not built say so, and tell the agent to record the gap instead of guessing.
URL pattern: /use-cases/<id>/openspec.md, all plans at /use-cases/index.json. Checkboxes on this page are for reading only.
Why
A launch post says what shipped, but nobody checked that the docs match the release, and a claim made on launch day cannot be checked a month later. Buyers, press and investors repeat launch claims. A wrong one is expensive to chase down; a claim with a record behind it is one less question in every sales call.
What Changes
- Add the
launch-with-proofworkflow to the PromptReports workflow library, run by the workflow runner as an agent acting for a person. - ContextFile: Check each docs page against the release's context version (today: Built).
- RepoOps: Trace what shipped from the merged changes and their agent sessions (today: Planned).
- PromptReports: Draft the announcement, one claim per sentence (today: Planned).
- PromptCompute: A person approves the announcement before it publishes (today: Planned).
- AI Source of Truth: Record each claim with its source change (today: Built).
- PromptReports: Show the record badge on the post (today: Switching on).
- Gate
cms.post.publish,social.post.publish,aisot.record.publishthrough PromptCompute approvals; each gated step runs only afterverifyResumepasses, withfinal_args. - Write a signed record at the end of each run and count one workflow request.
Impact
- Products: ContextFile, RepoOps, PromptCompute, AI Source of Truth, PromptReports.
- Outside the family: GitHub, Your blog or CMS, A social account (optional).
- State: Partly built. Three of five parts merged. The RepoOps release trace and the PromptCompute approval are still to build.
- Plan row: LDG-2269.
1. Connect and authorize
- 1.1 Confirm the org has a connection for each of: ContextFile, RepoOps, PromptCompute, AI Source of Truth, PromptReports. Stop and tell the user which is missing.
- 1.2 Get a PromptReports ID agent token for each product the run calls (RFC 8693 exchange). Where a product does not accept ID tokens yet, use its connect key.
- 1.3 Register policy for
cms.post.publish,social.post.publish,aisot.record.publishwithrequire_approvalas the default.
2. Build the steps
- 2.1 ContextFile: Check each docs page against the release's context version.
- 2.2 RepoOps: Trace what shipped from the merged changes and their agent sessions. Blocked: this part is Planned; until it ships, record the step's evidence as
missingand continue. - 2.3 PromptReports: Draft the announcement, one claim per sentence. Blocked: this part is Planned; until it ships, record the step's evidence as
missingand continue. - 2.4 PromptCompute: A person approves the announcement before it publishes. Blocked: this part is Planned; until it ships, record the step's evidence as
missingand continue. - 2.5 AI Source of Truth: Record each claim with its source change.
- 2.6 PromptReports: Show the record badge on the post.
3. Verify and record
- 3.1 Run a dry run with the user's inputs: Release tag = v2.8.0; Docs to check = docs/, the changelog; Where to publish = Blog, LinkedIn.
- 3.2 Run it once for real; approve each gate as the user.
- 3.3 Measure the outcomes: Share of launch claims with a public record: target 100%; Docs pages out of date on launch day: target 0; Hours from release tag to approved announcement.
- 3.4 Confirm the run left a signed record with a task id and one workflow request counted.
ADDED Requirements
Requirement: Every claim carries a record
The workflow SHALL attach an AI Source of Truth record id to each claim in an approved announcement.
Scenario: Every claim carries a record (1)
- WHEN the announcement is approved
- THEN each claim sentence links to a record that names its source change
Scenario: Every claim carries a record (2)
- WHEN a claim has no source change
- THEN the step marks it missing and the approval card lists it before the person decides
Requirement: Nothing publishes without a person
The publish step SHALL run only with a verified PromptCompute resume credential for this run and step.
Scenario: Nothing publishes without a person (1)
- WHEN the publish step receives a decision without a resume credential
- THEN the step refuses with missing_credential and the run waits
Scenario: Nothing publishes without a person (2)
- WHEN the approver edits the text
- THEN the step publishes final_args, never the first draft
Requirement: Evidence stays honest
Each step SHALL report what it saw as measured, inferred or missing, and SHALL never report missing as measured.
Scenario: A product part is not built
- WHEN a step depends on a part that is Planned or Proposed
- THEN the step records
missing, names the part, and the approval card shows the gap
Requirement: One run is one workflow request
The runner SHALL count one workflow request when the first step starts, and SHALL NOT count retries, resumes or dry runs.
Scenario: A gated step resumes after two days
- WHEN a person approves a step and the run resumes
- THEN the org's usage does not change