mediate
Skill0 starsChanged 4 months agoMITView on GitHub
Ownership-aware gate for work that may belong to another thread/session. Decide in-bounds vs out-of-bounds; if in-bounds, allow; if out-of-bounds, do NOT halt-and-wait and do NOT just proceed - notify the current owner with full details and let them judge whether the work should be done. Use when about to start/redo/prep work, edit files, or run experiments another thread may own, or when the user says "/mediate", "who owns this", "is this in bounds".
In this repository13 other files it gives its agents.
- AGENTS.mdAGENTS.md
- fleet/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- .cursor/rules/atlas.mdcCursor rule
- add-repoSkill
- atlas-setupSkill
- councilSkill
- onboardSkill
- optimizeSkill
- registry-syncSkill
- researchSkill
- snapshotSkill
- sync-repo-knowledgeSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Before doing work that may belong to another thread/session, decide in-bounds (this session owns it) vs out-of-bounds (another thread owns …
- Distinctive: line 26
Write what the owner needs to decide without re-deriving context:
- Distinctive: line 31
Do not proceed without explicit approval.
- Distinctive: line 35
If declined/deferred, leave the area untouched and note it.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
/mediate - ownership-aware work gating (route, don't halt)
Before doing work that may belong to another thread/session, decide in-bounds (this session owns it) vs out-of-bounds (another thread owns it). The point is to not block on the human and not silently overstep - for out-of-bounds work, notify the owner with enough detail to decide.
Step 1 - Identify ownership
Consult, in order: OWNERSHIP.md → any session-rename declaration ("this session owns X") →
repo/dir conventions (experiment dirs, manuscript, shared surfaces). Classify the target work as
IN-BOUNDS, SHARED/UNOWNED, or OUT-OF-BOUNDS. If genuinely indeterminate → treat as
OUT-OF-BOUNDS.
Step 2 - Route
- IN-BOUNDS → allow; proceed with normal rigor.
- SHARED/UNOWNED → read-only / low-risk additive edits fine; anything destructive or that another session is building on → treat as OUT-OF-BOUNDS.
- OUT-OF-BOUNDS → do not do the work. File a mediation request (Step 3); continue with in-bounds work meanwhile (don't idle).
Step 3 - Mediation request (out-of-bounds)
Write what the owner needs to decide without re-deriving context:
- What (specific change/run, exact paths) · Why · The change (diff/command/steps so "yes"
is one action) · Blast radius / reversibility · Recommended verdict (your honest call + reason).
Route it to the owner's channel (a decisions inbox if you have one, else a
MEDIATION/REQ-<date>-<slug>.mdnote + a notification). Owner unknown/unreachable → the human maintainer is the fallback owner. Do not proceed without explicit approval.
Step 4 - Record
Log the verdict (owner + decision + date). If approved, do the work and attribute it to the request. If declined/deferred, leave the area untouched and note it.
Contract
In-bounds → allow. Out-of-bounds → notify the owner with what/why/diff/risk/recommendation and let them judge. Never halt-and-wait on the human for an owner's call; never silently overstep.
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.

