logdet-postmortemlisted
Install: claude install-skill vivekbhange-hub/log-detective-skill
# logdet-postmortem: the write-up that only says what the logs show
A post-mortem is useful when readers can trust it. This skill writes one from the outputs of
`logdet-timeline`, `logdet-triage` and `logdet-trace`, cites a source for every factual claim,
and says plainly where the evidence stops. It has no script of its own: it works on those files.
## Inputs
- `logs-triage/timeline.md` (from `logdet-timeline`); required.
- `logs-triage/clusters.md` and `logs-triage/traces.md` if available.
- From the user, not from the logs: customer or business impact figures, who was on call, what
was done to stop it and when, any decisions. Ask; never guess.
- Commands below say `python3`; on Windows use `python` (or `py -3`) if a step needs the
other skills' scripts.
## Outputs
- `docs/postmortems/YYYY-MM-DD-<slug>.md` (or the user's location), using the structure below.
- A list in chat of what you could not establish and what would settle it.
## Steps
1. **Confirm authorization and audience.** The logs must be the user's to analyse, and the
document will be read by people who may not have access to them: quote nothing sensitive.
If a timeline or cluster file is missing, run the matching skill first; do not reconstruct
from memory.
2. **Gather the facts** from the files into a private list: first bad line, spike window, new
signatures, trace groups, the moment errors stopped. Give each a citation tag such as
`[timeline: db:18]`, `[cluster #3]`, `[trace #1]` o