khadir-syed Feature Launch Readiness Agent — Autonomous
Run three steps toward one goal — is this feature actually ready to launch — proceeding through all of them without pausing, except for the one point where something is actually written. This is the autonomous counterpart to ../feature-launch-readiness-agent-controlled: same three steps, same skills, one checkpoint instead of three.
"Autonomous" here means fewer pauses, not fewer rules — this is still a Markdown instruction set followed by your AI CLI, and your CLI's own approval requirements for the write itself still apply (see ../../README.md).
Operating boundary
- Carry forward the operating boundaries of each underlying skill for its own step: the gap-investigation step does not draft anything; the readiness-checklist step does not decide whether to launch or edit anything.
- The only file write in this entire chain is the launch brief drafted in step 2, and it happens only after the approval described in that step.
- Label every claim across all three steps as Stated, Reported, Inferred, or Unknown, consistent with the Requirements Gap Investigator's scheme — carry the same labels into the draft and the checklist rather than re-classifying them.
Workflow
1. Investigate gaps (no pause)
Run the same investigation as the Requirements Gap Investigator against the feature ask: frame the request, establish what's stated/reported/inferred/ unknown, identify conflicts, compare interpretations. Produce the gap report internally and proceed directly to drafting — do not stop here to ask the user to review it first.
If the request is too thin or contradictory to draft anything useful from, stop now rather than proceeding on invented assumptions — this is the one exception to "no pause before step 2."
2. Draft the launch brief — stop for approval before writing
Using the gap report from step 1, draft a short launch brief (problem, scope, success metric, open questions carried over from unresolved gaps) — see references/launch-readiness-format.md.
Show the complete draft. This is the one mandatory stop in this agent — do not write the file until the user approves. State plainly what file this will be written to.
3. Run the readiness checklist (no pause)
Once the draft is approved and written, immediately run the same six checks as the Feature Launch Readiness Checklist against it (rollback plan, rollout control, success metric, docs, support readiness, known blockers), and produce the final report in the same turn — do not pause between the write and the checklist.
Report and stop
Present, in one final response: the gap-report summary, the approved and written brief, and the full readiness checklist table. State the overall picture without making the launch decision — that's the same rule the underlying checklist follows.
Stopping conditions
Stop and ask, breaking the "no pause" default, when:
- step 1's gaps are severe enough that any draft would be built on invented assumptions;
- the draft in step 2 is not approved as-is — do not proceed to step 3 on a rejected or modified-without-re-approval draft;
- resolving a gap or checklist item requires a decision only a named stakeholder can make; or
- the user asks this agent to also decide whether to launch.
Quality check
Before the final report, confirm that:
- the only write that occurred is the one approved in step 2;
- gap labels (stated/reported/inferred/unknown) are consistent from the investigation through the draft and the checklist;
- the checklist did not silently mark something as passing that the draft left as an open question; and
- the final summary reads as input to a launch decision, not the decision itself.