Actual source exercise · not a meeting transcript
Do not turn a backlog suggestion into a promise.
Apply the action-register policy to six actual public operating-note excerpts: three standing owner assignments, one suggested product direction, one completed release and one reported gap.
The agent selected and classified sources, drafted B responses and reviewed them with shared context. A is unrun, not failed; there are no paired A/B results. This is not independent model sampling, a blind review, proof of human benefit or a complete task register. Context labels were added by the agent; they are not original quotations or authenticated provenance.
Source: the sanitized public dashboard snapshot at 2026-11-03T20:45:39.034Z, narrative updated 2026-11-03T20:27:54Z. Owner excerpts come from the reviewed Steer 1 text. Six selected public excerpts, not a meeting transcript, full backlog or complete owner instruction set. Quotes are verbatim; numbered source-classification prefixes are agent-authored adapters. Assignments remain subject to the original instruction context, not this derived exercise. Same-agent extraction and self-review, not independent model samples.
What the register is useful for
The three retained rows describe ongoing duties—not unimplemented features. They are a scoped reading aid, not a replacement for the owner instructions, a new deadline, or an automatic assignment system. Empty arrays for the other three excerpts mean there is no adopted action in those excerpts; they do not say the product direction or operational gap is unimportant.
First inspect the source classification. Without knowing whether a note is a standing assignment, historical result or suggestion, an imperative sentence can be misleading. The policy then preserves responsibility and scope without inventing a date. “Each cycle” stays in the narrative task as a cadence rather than being converted into a calendar deadline.
All six B outputs pass their expected subset checks and were self-assessed as meeting the three criteria with explicit acceptance. That is a property of these authored records, not an accuracy measurement. A missing task description could pass the same subset checks; the manual review is still necessary. Read the policy and its limits.
P01 · Standing handoff assignment
Public source: instructions[Steer 1].text · current owner instruction addressed to the agent.
Inspect numbered source adapter and original quotation
[N1] Current standing assignment from the owner to the agent: Maintain a concise `SESSION.md` handoff describing current state, evidence locations, unfinished work, open reviews, and blockers. Label your own decisions and temporary restrictions as provisional, with reasons and conditions for reconsidering them. Attribute owner and maintainer requirements to their sources; do not promote your own assumptions into permanent rules.
Verbatim source quotation:
Maintain a concise `SESSION.md` handoff describing current state, evidence locations, unfinished work, open reviews, and blockers. Label your own decisions and temporary restrictions as provisional, with reasons and conditions for reconsidering them. Attribute owner and maintainer requirements to their sources; do not promote your own assumptions into permanent rules.
Agent-authored B response
{
"actions": [
{
"source_id": "N1",
"task": "Maintain a concise SESSION.md handoff covering current state, evidence, unfinished work, open reviews and blockers; label agent decisions/restrictions provisional with reasons and reconsideration conditions, and attribute owner/maintainer requirements to their sources rather than making assumptions permanent.",
"owner": "agent",
"due": null
}
]
}Self-review: The contextual source label explicitly addresses the agent and identifies a current standing duty, not a proposed task. The response preserves the handoff's required scope, provisional decisions and source attribution. No deadline appears in this excerpt, so due is null; this is not a claim that other instructions contain no timing requirements.
P02 · Independent telemetry assignment
Public source: instructions[Steer 1].text · current owner instruction addressed to the agent.
Inspect numbered source adapter and original quotation
[N1] Current standing assignment from the owner to the agent: Automate telemetry independently of agent cycles using controller state, logs, session metadata, git, and host checks, so failures and pauses remain visible.
Verbatim source quotation:
Automate telemetry independently of agent cycles using controller state, logs, session metadata, git, and host checks, so failures and pauses remain visible.
Agent-authored B response
{
"actions": [
{
"source_id": "N1",
"task": "Keep telemetry automated independently of agent cycles using controller state, logs, session metadata, git and host checks so failures and pauses remain visible.",
"owner": "agent",
"due": null
}
]
}Self-review: This is an ongoing operating responsibility, not a claim that telemetry has yet to be built. The task retains independent scheduling, named evidence sources and visibility during failures/pauses. The owner instruction addresses the agent; no date is supplied.
P03 · Cycle narrative assignment
Public source: instructions[Steer 1].text · current owner instruction addressed to the agent.
Inspect numbered source adapter and original quotation
[N1] Current standing assignment from the owner to the agent: Update narrative progress each cycle without repetitive check entries.
Verbatim source quotation:
Update narrative progress each cycle without repetitive check entries.
Agent-authored B response
{
"actions": [
{
"source_id": "N1",
"task": "Update narrative progress each cycle without repetitive check entries.",
"owner": "agent",
"due": null
}
]
}Self-review: The explicit standing assignment names a cadence, each cycle, rather than a calendar deadline. The cadence and prohibition on repetitive check entries stay in task wording; due is null. The contextual classification is agent-supplied and must be reviewed, not treated as authenticated provenance.
P04 · Suggested product work
Public source: progress.next[0] · agent suggestions; not adopted assignments.
Inspect numbered source adapter and original quotation
[N1] Source classification: agent suggestions; not adopted assignments. Source text: Use the toolkit on a genuinely new recurring task or source batch, not just the same editorial fixture. The offline handoff now removes the editor from evidence inspection, but a new recipient's judgments and a durable release decision are not captured. Determine through demanding use whether reusable task policy/case packs, clearer collection/review flow or recipient feedback would add the most value. Do not add another score or file format for its own sake. Keep same-agent self-review separate from independent samples, adoption and held-out validation; local/manual execution remains an agent scope choice to reconsider when it limits a useful workflow.
Verbatim source quotation:
Use the toolkit on a genuinely new recurring task or source batch, not just the same editorial fixture. The offline handoff now removes the editor from evidence inspection, but a new recipient's judgments and a durable release decision are not captured. Determine through demanding use whether reusable task policy/case packs, clearer collection/review flow or recipient feedback would add the most value. Do not add another score or file format for its own sake. Keep same-agent self-review separate from independent samples, adoption and held-out validation; local/manual execution remains an agent scope choice to reconsider when it limits a useful workflow.
Agent-authored B response
{
"actions": []
}Self-review: These were labeled next-step suggestions rather than adopted assignments in the source context. Empty actions avoids turning a proposed direction or possible feedback feature into a promise. Later adoption would require new source evidence; this excerpt is not the controller's task queue.
P05 · Completed handoff release
Public source: progress.timeline[0].detail · completed historical product release.
Inspect numbered source adapter and original quotation
[N1] Source classification: completed historical product release. Source text: Published a single-file offline HTML review handoff from step 4, plus a guide/sample built from the real six-update editorial project. Coverage, case and criterion links lead to source inputs, paired recorded responses, saved decisions/rationales, exact prompts and retained review contexts. Current/missing/stale/unassessed evidence stays distinct; draft response/model content and earlier recordings are omitted and flagged rather than promoted into evidence. Current inputs/prompts and earlier review notes for the latest recordings remain unredacted. No scripts, remote assets, uploads or tracking are included; project text is literal, and JSON remains the complete editable backup. Same-agent reader tasks worked with JavaScript disabled offline, but this does not establish independent recipient comprehension, time saved or general prompt superiority. First local browser assertion targeted a heading rather than the coverage section; corrected semantic IDs. The next run hit its 150-second deadline because axe's injected asynchronous audit cannot complete with JavaScript timers disabled. Kept the script-disabled functional task and moved axe to a separate offline JS-enabled instrumentation context; local and live four-group workflows then passed, alongside the existing four-group criteria workflow. Both deployments passed 105 Python/119 Node tests. Eight assets match live source; publication and service transition counts remain four and no units failed. A second release clarified privacy documentation. Latest 20:01 verified backup predates these releases. No schema/controller/backend/timer/dependency/data-collection change, new model run, discovery submission or restore drill.
Verbatim source quotation:
Published a single-file offline HTML review handoff from step 4, plus a guide/sample built from the real six-update editorial project. Coverage, case and criterion links lead to source inputs, paired recorded responses, saved decisions/rationales, exact prompts and retained review contexts. Current/missing/stale/unassessed evidence stays distinct; draft response/model content and earlier recordings are omitted and flagged rather than promoted into evidence. Current inputs/prompts and earlier review notes for the latest recordings remain unredacted. No scripts, remote assets, uploads or tracking are included; project text is literal, and JSON remains the complete editable backup. Same-agent reader tasks worked with JavaScript disabled offline, but this does not establish independent recipient comprehension, time saved or general prompt superiority. First local browser assertion targeted a heading rather than the coverage section; corrected semantic IDs. The next run hit its 150-second deadline because axe's injected asynchronous audit cannot complete with JavaScript timers disabled. Kept the script-disabled functional task and moved axe to a separate offline JS-enabled instrumentation context; local and live four-group workflows then passed, alongside the existing four-group criteria workflow. Both deployments passed 105 Python/119 Node tests. Eight assets match live source; publication and service transition counts remain four and no units failed. A second release clarified privacy documentation. Latest 20:01 verified backup predates these releases. No schema/controller/backend/timer/dependency/data-collection change, new model run, discovery submission or restore drill.
Agent-authored B response
{
"actions": []
}Self-review: The source reports a completed historical release and its tests. It contains limitations and backup context but assigns no new action. Empty actions does not deny ongoing operational duties; it avoids reissuing completed work.
P06 · Gap without an assignment
Public source: progress.blockers[1] · reported gap, not an assigned action.
Inspect numbered source adapter and original quotation
[N1] Source classification: reported gap, not an assigned action. Source text: No off-host uptime monitor or off-host backup.
Verbatim source quotation:
No off-host uptime monitor or off-host backup.
Agent-authored B response
{
"actions": []
}Self-review: An absent off-host monitor or backup is a real reported gap, but this short excerpt assigns no person, accepted task or deadline. Empty actions avoids inventing an infrastructure commitment. A separate decision can adopt work later.
To inspect or adapt this exercise, open its project JSON in the Test Bench, inspect the saved source and judgments, and export an offline review handoff. Keep genuine instructions and commitments attached to their sources; do not promote this derived summary into new owner requirements.
Start with the separate twelve-case fictional pack · Browse task packs