wawa
Skill8 starsChanged 56 days agoMITView on GitHub
Where are we at, and what's next — verified session/repo status with a recommended next move.
In this repository3 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- council-reviewSkill
- init-repo-docsSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 6 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 14
If it names a repo, area, or task, narrow to that.
- Distinctive: line 34
If this session already did work, what of it is actually on disk versus still only described in conversation.
- Distinctive: line 36
Check anything cheap that would change my next decision: does the build/test state still hold, is a published artifact current, does the …
- Very specific: line 42
Read STATUS.md and TODO.md (and DECISIONS.md if present) for intent and history.
- Distinctive: line 54
Keep it short and decision-shaped.
- Distinctive: line 72
If something is genuinely unknown, say "unknown" and name the one check that would resolve it.
- Distinctive: line 75
Do not overcomplicate language, checkups or findings.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
User-invoked. Run only when the user asked for this skill. Scope hint: $ARGUMENTS
"wawa" = where are we at now, and what's next.
Give me an orientation briefing. Read-only: never edit, commit, deploy, or start long jobs. If Scope hint is empty, cover the current session and the current repo. If it names a repo, area, or task, narrow to that.
Rule that overrides everything else here
Do not tell me the state of things by quoting notes. STATUS.md, TODO.md,
planning docs, and anything I said earlier in this session are hypotheses with a
timestamp. Check what is actually true right now, then report. If a note and
reality disagree, the disagreement IS the finding and goes near the top.
Label every claim verified: <what you ran> or assumed. An unlabeled claim reads
as verified, which is the failure mode this command exists to prevent.
Step 1: Establish real state (run these, don't reason about them)
git status --shortandgit log --oneline -8. What is uncommitted, what landed recently, which branch, ahead/behind.- Any work in flight I started: background tasks, running processes, spawned agents, open monitors. For each: still alive, exited, or silently dead? A process alive but producing no output is running, not finished — say which, and say how long it has been going. Never report a pending job as done, and never invent its result.
- If this session already did work, what of it is actually on disk versus still only described in conversation.
- Check anything cheap that would change my next decision: does the build/test state still hold, is a published artifact current, does the thing I think is deployed match what is live.
Step 2: Read the tracking files, then distrust them
Read STATUS.md and TODO.md (and DECISIONS.md if present) for intent and
history. For a planning, strategy, or "what should we do" scope, also read any
project wiki or history page the repo points at (README, CLAUDE.md, or a local
wiki path). Skip if none exists. Do not propose building something that already
exists.
Then spot-check the load-bearing claims. Items marked done that are not done, items marked blocked whose blocker has cleared, and dates that have passed are the most valuable things you can surface.
Step 3: Report
Keep it short and decision-shaped. Findings and decisions, not narration. No recap of what I already watched you do.
Where we are — 3-6 bullets. Current state of the work, each one verified.
In flight — anything running or half-finished, with its real status and what it is waiting on. Say plainly if something needs me (a permission, a credential, a manual step, a decision only I can make).
Drifted — where the notes, the plan, or my assumptions no longer match reality. Omit the heading if nothing drifted. Include overdue dated items.
What's next — a short ordered list. For each: the move, why it is next, and roughly what it costs. Mark anything blocked on me.
Recommendation — one line: the single next action you would take, and whether you need my go-ahead to take it.
If something is genuinely unknown, say "unknown" and name the one check that would resolve it. Do not fill the gap with a plausible guess.
Do not overcomplicate language, checkups or findings. Aim for simplicity and clarity.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

