rollback-postmortem

rollback-postmortem is a skill for Claude Code, Codex from iampawan/Manifest. It costs 85 tokens per session (1,910 once invoked), scanned A, original, MIT.

A blameless report created after a feature has been rolled back, meaning the change was disabled or the previous release was restored. It reconstructs the event timeline, records the result, posts the report to Slack, and reopens follow-up work.

In plain words
What is it for?
Use it after a human confirms a rollback to review the contract timeline, guard reports, and Sentry data, then record what happened and what should be done next.
Why use it?
It turns a rollback into a documented learning process instead of ending with only a change in production. It also prevents reports from being written before a rollback is confirmed.

Skill for Claude CodeCodex

Part of the manifest plugin — 28 skills, 15 commands shipped together

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/iampawan/manifest/rollback-postmortem
Any agent
npx skills add iampawan/Manifest --skill rollback-postmortem
Clone the repo
git clone --depth 1 https://github.com/iampawan/Manifest

Made for: Claude Code, Codex.

Or install manifest, the plugin that ships this one along with the rest of its 28 skills, 15 commands.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for rollback-postmortem

README.md
[![agentmods](https://agentmods.dev/badge/skills/iampawan/manifest/rollback-postmortem.svg)](https://agentmods.dev/skills/iampawan/manifest/rollback-postmortem)
Your own site
<a href="https://agentmods.dev/skills/iampawan/manifest/rollback-postmortem"><img src="https://agentmods.dev/badge/skills/iampawan/manifest/rollback-postmortem.svg" alt="Measured on agentmods" height="20"></a>
Per session 85 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 1,910 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00085 $0.01910
Opus 5 $0.00043 $0.00955
Sonnet 5 $0.00017 $0.00382
Haiku 4.5 $0.00009 $0.00191

Measured 3d ago against content hash e76353e73481, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

rollback-postmortem scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/rollback-postmortem/SKILL.md · 192 lines

How it starts

The opening of the file, as written. The whole thing — 192 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Rollback postmortem

When rollback-guard recommends a rollback and the team reverts, the loop has to actually end somewhere — with the status recorded and the lesson captured, not just a flag flipped and silence. That's this skill.

It runs after a human has confirmed the rollback happened (you don't flip flags — see rollback-guard). It's blameless by construction: the output describes what the system did and what was learned, never who to blame.

Inputs

  • Contract ID
  • Optional: a one-line reason from the human ("iOS crash on cold start")
  • Optional: the guard report that triggered it (.manifest/contracts/<ID>.guard-*.md)

Process

1. Confirm a rollback actually happened

Read the contract. A postmortem is only appropriate once the feature is no longer at its rollout stage (flag flipped off / previous release redeployed). If the latest guardVerdict isn't recommend-rollback and the human gave no reason, ask the human to confirm the rollback before proceeding — don't write a postmortem for a feature that's still live.

2. Reconstruct the timeline (from data the contract already has)

You don't need new infrastructure — the contract's own cycle-time stamps plus the guard/deploy reports are the timeline:

  • promotedAt, prMergedAt, qaDeployedAt, canaryStartedAt, currentRolloutPercent — when each stage happened and how far it got.
  • The guard-*.md reports — the signal that breached, when it was first seen, and the values vs budget.
  • Sentry (via MCP) — the top issues for the release tag: error class, first-seen, affected users, stack location.
  • The deploy report(s) — what verification said before rollout.

3. Write the postmortem

.manifest/contracts/<ID>.postmortem.md:

---
contractId: <ID>
revision: <N>
type: rollback-postmortem
generatedAt: <ISO>
rolledBackAt: <ISO>
rolledBackAtStage: 10%
---

# <title> — rollback postmortem

## What happened
One paragraph: what shipped, how far it ramped, what signal breached,
when it was caught, and what action was taken. Plain, blameless.

## Timeline (IST / UTC)
| When | Event |
|---|---|
| <t> | Promoted; SLA started |
| <t> | Merged; deployed to QA (verify: ready-for-canary) |
| <t> | Canary started at 1% |
| <t> | Advanced to 10% |
| <t> | rollback-guard: recommend-rollback (error rate 4.2× baseline) |
| <t> | Flag flipped to 0% by <owner> |

## Signal that breached
The guard verdict + the signal table: observed vs budget, adoption %,
and the top Sentry issue(s) with first-seen and affected users.

## Contributing factors
Blameless. What about the change, the spec, or the verification let this
reach users? Was there an edge case the contract didn't cover (link the
finding if so)? Did verification pass because the failure only appears
under real traffic / a cohort QA didn't exercise?

## What went well
Detection worked — the guard caught it at N% before full rollout. Note
the blast radius that was avoided.

## Action items
- [ ] <owner> — concrete fix (with a tracking ref)
- [ ] Add the missed edge case to the contract before re-promoting
- [ ] (if verification gap) strengthen the AC / device-cohort coverage

## Re-ship plan
The contract reopens at the spec phase: address the contributing
factors in a new revision, re-verify, re-promote. Link the follow-up.

Read the full file on GitHub · 192 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 3d ago First seen · 192 lines · 85 tokens per session scan A e76353e73481

Subscribe to this mod's changes

rollback-postmortem is a skill published in the GitHub repository iampawan/Manifest (5 stars, last pushed yesterday), licensed MIT. It adds 85 tokens to every session and 1,910 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.