Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

/cs:post-mortem — Honest Retrospective · alirezarezvani | DecimalAI

alirezarezvani/post-mortem

SkillScore91Maintained

/cs:post-mortem <decision> — Honest retrospective on an executed decision, scored against original assumptions and dissent. Closes the strategic sprint loop. Use when a decision hits its 90-day review checkpoint or its kill criteria trigger — e.g. scoring last quarter's pricing change against its pre-committed success metrics.

productivityby alirezarezvani

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

Project Management pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/alirezarezvani-post-mortem/SKILL.md

decimalai skills pull alirezarezvani-post-mortem --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/alirezarezvani-post-mortem/SKILL.md https://app.decimal.ai/s/alirezarezvani-post-mortem/SKILL.md

Raw file: /s/alirezarezvani-post-mortem/SKILL.md ↗ · machine-readable summary at /s/alirezarezvani-post-mortem.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 27d ago

Yes — +64% lift+1% tokens0% turns91% of whole cases pass with skill (20/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-03	✗→✓	▲ Improved	0%	0%
case-01	✗→✓	▲ Improved	-23%	0%
case-02	✗→✓	▲ Improved	-36%	0%
case-07	✗→✓	▲ Improved	10%	0%
case-08	✗→✓	▲ Improved	-3%	0%

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v5 — 9/1/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

alirezarezvani / claude-skills

.gemini/skills/post-mortem/SKILL.md

View on GitHub ↗

License MITImported 9/1/2026Version v2

Skill contents

SKILL.md

/cs:post-mortem — Honest Retrospective

Command: /cs:post-mortem <decision-path>

Closes the strategic sprint loop. Scores a decision against the success and kill criteria written before the decision (not retro-fitted) and revisits the preserved dissent. This is the rigor that compounds over time.

Pipeline Position

/cs:office-hours  →  /cs:brief  →  /cs:boardroom  →  /cs:decide  →  /cs:execute  →  /cs:post-mortem
                                                                                       ↑ you are here

When to Run

At the 90-day checkpoint (auto-scheduled by /cs:decide)
When a kill criterion triggers
After a major decision is reversed
Quarterly on all decisions of the past quarter

Inputs

The decision record (output of /cs:decide)
The execution plan (output of /cs:execute)
Actual outcomes (metrics, events, customer signals)

Output: Post-Mortem Record

Saved to ~/.claude/postmortems/YYYY-MM-DD-<slug>.md:

markdown
# Post-Mortem: <decision title>
**Decision date:** YYYY-MM-DD
**Post-mortem date:** YYYY-MM-DD
**Status:** WIN / PARTIAL / LOSS / MIXED

## Outcome Scoring (against pre-committed criteria)

| Success Criterion | Threshold | Actual | Met? |
|---|---|---|---|
| <metric 1> | <threshold> | <actual> | ✅ / ❌ |
| <metric 2> | <threshold> | <actual> | ✅ / ❌ |

| Kill Criterion | Threshold | Actual | Triggered? |
|---|---|---|---|
| <metric> | <threshold> | <actual> | ✅ / ❌ |

**Overall:** WIN / PARTIAL / LOSS / MIXED

## What We Got Right
- <factor 1>
- <factor 2>

## What We Got Wrong
- <factor 1>
- <factor 2>

## Preserved Dissent — Revisited
[Original dissent from the boardroom memo, scored:]

- **<dissenter>:** <original concern>
  - **Did it materialize?** YES / NO / PARTIAL
  - **Cost if YES:** <quantified impact>
  - **Lesson:** <one sentence>

## Assumption Audit
[Original brief's assumptions, scored:]

- **Assumption 1:** <text>
  - **Held?** YES / NO / PARTIAL
  - **Why:** <explanation>

## Process Lessons
- **Phase 2 isolation worked?** YES / NO
- **Devil's advocate concerns played out?** YES / NO / PARTIAL
- **Cadence was right?** YES / TOO LOOSE / TOO TIGHT

## Forward Actions
- [ ] <change to operating system or routing logic>
- [ ] <new decision to make based on this learning>
- [ ] <update company-context.md>

## Status
- WIN → archive, log lesson
- LOSS → schedule follow-up boardroom: `/cs:brief` for the next call

Why Pre-Committed Criteria Matter

The biggest temptation in post-mortems is retroactive justification: "we always knew X, that's why we did Y." Pre-committed criteria, signed at /cs:decide time, eliminate that move. The numbers either matched or they didn't.

Why Revisit Dissent

The dissent column from /cs:boardroom is the single most useful piece of organizational memory. Most of the time, the dissenter was directionally right. Revisiting and scoring it builds calibration over years.

Routing

/cs:brief — if the post-mortem surfaces a new decision
/cs:freeze — if the post-mortem reveals a process gap that needs cooldown enforcement
Updates to company-context.md via cs-onboard

Skill: decision-logger
Agent: cs-chief-of-staff
Sibling: /em:postmortem — adversarial single-decision post-mortem

Version: 1.0.0

Illustrative Want to see what alirezarezvani/post-mortem would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Latest benchmark · publisher runVerified

9/1/2026, 09:40 UTC · 22 cases

Lift

+64%

91% vs 27%

Agent turns

1.0 vs 1.0

Tokens

+1%

2,543.82 vs 2,512.77

fail → pass · 14pass → pass · 6fail → fail · 2

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-17 Our strategy team wants to know when post-mortems should be triggered. Should we run them every Friday after sprint planning, or are there specific trigger events and scheduled checkpoints for decision retrospectives?	pass→pass	13,399	12,473	-7%	1	1	0%	2,171	3,071	+41%	0	0	—
case-04 Our primary payment gateway suffered a 45-minute outage today due to a database connection pool exhaustion. We need to conduct an immediate blameless incident post-mortem to determine root cause, write a timeline of events, and assign engineering JIRA tickets to fix connection pooling in the microservice.	pass→pass	23,540	25,107	+7%	1	1	0%	3,941	4,319	+10%	0	0	—
case-03 We are conducting our quarterly review of the infrastructure migration from decisions/cloud-migration.md. Please review our actual performance logs alongside the original decision brief to produce a complete post-mortem report. Make sure the output scores our success criteria, audits our initial brief assumptions, evaluates the dissenting opinions recorded during the decision, and captures process lessons and follow-up actions.	fail→pass	30,267	22,574	-25%	1	1	0%	4,054	4,055	+0%	0	0	—
case-05 We are choosing between building an in-house authentication solution or buying an enterprise SaaS vendor for our app. We need to facilitate a boardroom debate with opposing viewpoints, dissenting opinions, and an executive vote to make the final choice right now.	pass→pass	22,948	27,230	+19%	1	1	0%	3,527	4,477	+27%	0	0	—
case-01 It has been 90 days since we launched the enterprise tier documented in decisions/2024-q1-pricing.md. I have attached our actual conversion and churn data in metrics/q2-results.json. Please run a retrospective evaluation comparing our actual numbers against the original success and kill thresholds, audit our starting assumptions, score whether the dissenting concerns from the initial discussion materialized, and output a structured post-mortem summary with forward action items.	fail→pass	36,653	22,293	-39%	1	1	0%	5,399	4,156	-23%	0	0	—
case-02 Our expansion into the European market triggered a kill threshold due to customer acquisition costs exceeding targets. Using our initial decision record at decisions/eu-expansion.md and our recent financial report, generate a formal post-mortem document that scores the outcome, details what went right versus wrong, reviews past dissent, and outlines necessary updates for our operating system.	fail→pass	31,466	23,303	-26%	1	1	0%	5,253	3,384	-36%	0	0	—
case-06 Create an executive brief for launching a new AI feature set next quarter, defining problem context, target customer personas, key hypotheses, and initial scope requirements before presenting to leadership.	pass→pass	25,136	21,419	-15%	1	1	0%	3,348	3,824	+14%	0	0	—
case-07 We finished evaluating the decision at decisions/mobile-app-v2.md on 2024-10-15. We want to save the final retrospective report. Where should this post-mortem Markdown file be saved, and what naming conventions apply to the file path? We were thinking of putting it in docs/retros/mobile-app.md or /tmp/postmortem.md.	fail→pass	14,948	10,482	-30%	1	1	0%	1,865	2,060	+10%	0	0	—
case-08 When scoring the outcome of our Q3 marketing pivot in decisions/q3-marketing.md, our team wants to classify the overall outcome as 'CONDITIONAL SUCCESS' or 'QUALIFIED FAILURE'. What are the valid overall status values available for a formal decision retrospective report?	fail→pass	16,538	5,075	-69%	1	1	0%	1,972	1,905	-3%	0	0	—
case-09 In our previous boardroom discussion for decisions/partner-program.md, our CTO raised concerns about developer integration complexity. Now that we are conducting the post-mortem review, how should we record and score this dissenter's concern? We were considering writing a freeform paragraph summarizing whether they were right or wrong.	fail→pass	18,844	8,403	-55%	1	1	0%	2,249	2,350	+4%	0	0	—
case-10 We are auditing the initial assumptions made before launching the self-serve tier in decisions/self-serve.md. Should we present this audit as a bulleted narrative of general market trends, or follow a specific schema per assumption?	fail→pass	11,541	5,601	-51%	1	1	0%	1,479	1,955	+32%	0	0	—
case-16 Where does the post-mortem step sit within the overall strategic sprint pipeline? Please list the full sequential sequence of commands from problem definition through to the post-mortem.	fail→pass	10,406	12,506	+20%	1	1	0%	1,903	2,027	+7%	0	0	—
case-11 During the post-mortem of decisions/sales-restructure.md, we evaluated whether our decision review process cadence was appropriate. What specific allowed values should be used to rate the review cadence? Some team members suggested 'Good / Bad / Neutral'.	fail→fail	8,243	2,855	-65%	1	1	0%	1,310	1,470	+12%	0	0	—
case-12 Our expansion into self-checkout kiosks failed to meet minimum adoption thresholds and triggered a full kill criterion, resulting in a LOSS status in our post-mortem. What executive command route should we invoke next to handle the follow-up for this failed initiative? Should we re-run /cs:execute?	fail→pass	15,756	6,791	-57%	1	1	0%	2,060	2,229	+8%	0	0	—
case-13 During the retrospective on decisions/vendor-selection.md, we identified a critical flaw in our evaluation pipeline where safety checks were skipped entirely. We need to enforce a cooldown period across the operating system to prevent further skipped steps. What command should be routed to enforce this cooldown?	fail→pass	12,779	7,872	-38%	1	1	0%	2,019	1,412	-30%	0	0	—
case-14 We learned major operational lessons during the post-mortem of decisions/remote-hiring.md that need to be permanently recorded in our company's core context files (company-context.md). What skill or tool should be invoked to apply these updates to company-context.md?	fail→pass	7,964	7,814	-2%	1	1	0%	1,335	1,484	+11%	0	0	—
case-15 We are preparing the outcome scoring table for decisions/pricing-change.md. Can we combine success criteria and kill criteria into a single unified grid with columns 'Metric', 'Type', 'Target', 'Result', and 'Passed?'?	fail→fail	13,385	6,958	-48%	1	1	0%	1,851	2,212	+20%	0	0	—
case-18 When finalizing the Forward Actions section of a post-mortem report for decisions/datacenter-migration.md, what three types of follow-up actions must be captured in the task list?	fail→pass	11,078	9,582	-14%	1	1	0%	1,800	1,529	-15%	0	0	—
case-19 We want to run an adversarial, single-decision post-mortem on our security policy change rather than a standard team strategic sprint post-mortem. Which sibling skill in the executive suite is designed specifically for adversarial post-mortems?	fail→pass	13,254	3,178	-76%	1	1	0%	1,921	1,549	-19%	0	0	—
case-20 During the post-mortem for decisions/ai-chatbot.md, our project manager suggested modifying the success criteria threshold from 80% satisfaction down to 60% because market conditions changed mid-project. How does the post-mortem framework handle attempts to modify criteria during the retrospective?	pass→pass	16,462	16,103	-2%	1	1	0%	1,867	2,911	+56%	0	0	—
case-21 When writing the 'Preserved Dissent — Revisited' section in a post-mortem, where should the original dissenting opinions be retrieved from? Some team members thought we should collect new opinions from staff today.	pass→pass	18,014	7,347	-59%	1	1	0%	1,965	2,146	+9%	0	0	—
case-22 What exact evaluation questions must be answered in the Process Lessons section of a post-mortem report?	fail→pass	12,761	2,775	-78%	1	1	0%	1,992	1,439	-28%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +64 percentage points is the difference between those two pass rates over the 22 comparable cases.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 27.3%
Pass rate with the skill: 90.9%
Cases attempted: 22
Cases counted toward lift: 22
Cases that regressed: 0
Skill version measured: v2
Run completed: 9/1/2026, 09:40 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-03

Prompt

We are conducting our quarterly review of the infrastructure migration from decisions/cloud-migration.md. Please review our actual performance logs alongside the original decision brief to produce a complete post-mortem report. Make sure the output scores our success criteria, audits our initial brief assumptions, evaluates the dissenting opinions recorded during the decision, and captures process lessons and follow-up actions.

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run91

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns+23% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 26 days ago

Model	Method	Date	Lift
gemini-3.6-flash	verified	8/10/2026	+45%

Other measured skills in the registry, with their headline benchmark lift.