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.
npx agentmods add skills/ardurai/ardur-agent/write-postmortemnpx skills add ArdurAI/ardur-agent --skill write-postmortemgit clone --depth 1 https://github.com/ArdurAI/ardur-agentWhat 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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00021 | $0.01179 |
| Opus 5 | $0.00010 | $0.00589 |
| Sonnet 5 | $0.00004 | $0.00236 |
| Haiku 4.5 | $0.00002 | $0.00118 |
Grade A, and why
write-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 2d 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.
How it starts
The opening of the file, as written. The whole thing — 108 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Writing a postmortem
A postmortem exists to make the same failure less likely next time. It is not a status report, not a defense, and not a record of who was at fault. Its value is entirely in the action items it produces and whether they get done.
The single most important word is blameless. The moment a postmortem reads as "Alice deployed the bad change", people stop being honest, and you lose the information you need to actually fix the system. Humans operating a system as designed will occasionally make mistakes; if a single human error could take the system down, the system is the problem to fix, not the human.
Use the skeleton in @./template.md to draft. The guidance below is how to fill
it well.
1. Timeline — what happened, when
A factual, timestamped sequence. No interpretation yet.
- Start before the incident: the change or condition that set it up.
- Mark the key moments: when it began, when it was detected, when it was acknowledged, key diagnostic and mitigation steps, when it was resolved.
- Use absolute timestamps with a timezone. "About 20 minutes later" is useless to a future reader reconstructing the sequence.
- Write what was known at the time, not what you know now with hindsight. "The on-call saw queue depth rising and assumed a traffic spike" is honest and useful; "obviously it was the bad deploy" is hindsight and unfair.
Two derived numbers fall out of a good timeline and are worth stating: time-to-detect (incident start to detection) and time-to-resolve (detection to resolution). They tell you whether to invest in better alerting or faster mitigation.
2. Impact — who and how much, quantified
- User-facing effect: what could users not do, and for how long.
- Scope: how many users, which regions/tenants, what fraction of traffic.
- Quantify it: requests failed, revenue affected, SLA/error budget consumed, data lost or delayed. A number makes the cost concrete and justifies the follow-up work.
3. Contributing factors — not "the root cause"
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 2d ago First seen · 108 lines · 21 tokens per session scan A 7a1c33c8a38b
write-postmortem is a skill published in the GitHub repository ArdurAI/ardur-agent (2 stars, last pushed 5d ago), licensed Apache-2.0. It adds 21 tokens to every session and 1,179 once invoked, about $0.0001 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.
Other skills, from other repositories
analyze-agent-codebase
当用户希望依据源码、配置、schema 和测试，分析 Coding Agent、Agent 框架或多 Agent 系统的真实架构、执行流程、上下文、记忆、工具、权限或扩展点时使用。继续回答同一分析中的定向机制问题或整理专题文档时也使用。普通代码评审、实现或修复任务，以及不需要仓库证据的概念问答不使用。.
worktree
当用户明确要求使用 Git worktree，或需要为一项独立开发工作创建、查找、复用、交接或清理隔离工作目录时使用。普通只读任务、非 Git 仓库，以及当前工作无需独立分支或工作目录时不使用。.
member-studio
当用户希望创建新的 Rovai 队员，或继续调整、确认本次创建中尚未写入的队员名牌和头像方案时使用。普通成员资料咨询、编辑已创建队员，以及只设计角色或头像但不加入名册的任务不使用。.
campfire
当用户希望 Camp 中多位成员共同讨论、从不同角度分析、比较方案、评估利弊或讨论后形成建议时使用。主持人发起和继续整理讨论，成员在收到本次讨论任务时也使用。普通单人问题、无关发言、迟到补充和已经结束的讨论不使用。.
tasteful-ui
UI design and implementation for real product surfaces with taste-first critique, reference routing, project-specific design briefs, variation comparison, and implementation verification. Use when Codex should redesign, build, polish, or critique frontend UI by reading project context, exploring taste, routing through…
grill-duo-with-docs
当用户希望在 Camp 中通过持续追问和一位固定搭档的独立复核，澄清计划或设计，并同步维护已确认的领域词汇、当前权威文档或版本决策记录时使用。邀请者继续处理用户回答或当前搭档建议，成员收到文档版双人追问复核任务时也使用。普通单人问答、无需维护领域文档的追问、多人讨论、无关发言和已经结束的会话不使用。.