agentleFS
rikvanrielrikvanriel/code-hygiene/skills/artifact-handoff/SKILL.md

artifact-handoff

Skill2 starsChanged 56 days agoMITView on GitHub

Use when handing an artifact to someone, or trusting one you received — identity, provenance, verification.

The file

Files
SKILL.md

Artifact Handoff

Capability: an artifact that leaves your hands can be identified, traced to its inputs, and trusted by whoever receives it — and an artifact you receive can be checked before you act on it.

Trigger

Producing a file for a person or another system (report, render, dataset, package, index, cache); receiving one; writing a derived value that will be read later.

Rules

GC-80 — A handed-over artifact identifies itself

The filename or an on-artifact label carries the source it came from, the run or date, and the parameters that shaped it. A label burned into the artifact survives a chat client that hides filenames; a directory listing does not. When several versions of the same artifact exist, the identifier is what keeps "the one I sent" from being a guess — for anyone, including you tomorrow.

GC-81 — A generated artifact retains the inputs it was derived from

A digest, summary, report or generated file records which source revision, size, or window it was computed from, so its claims stay re-verifiable after the logs rotate and the intermediate files are cleaned. A generated claim whose inputs are gone is indistinguishable from an invented one.

GC-82 — A cached or derived value records its provenance and invalidates on change

Anything cached, memoized, precomputed or otherwise derived carries the identity of what it was computed from (source version, file size and mtime, a hash, a parameter set) and is recomputed when that identity changes. Serving a cached result built from a different input is the quiet failure: the value looks fresh and is wrong.

GC-83 — Read back what you wrote

After writing to a store — memory, index, database, cache, file — read the value back from that store to confirm it is retrievable and correct. A successful write call is not evidence the value is there: dropped writes, silent schema mismatches and stale copies are all invisible until something reads.

GC-84 — A local success is not a remote state

A successful push is not a merged change; a submitted form is not an accepted value; a locally-correct file is not a deployed one. Confirm the state you actually care about with a fresh read of the authoritative source, and report the distinction between "I sent it" and "it landed".

Verification

# Artifact identity: name carries source + run, and the hash of what you send
sha256sum dist/*.tar.gz && ls -l dist/ | head
# Cache provenance: the key includes the input identity, not just the request
grep -rn "mtime\|sha256\|version" cache/ --include='*.py' | head -n 5
# Read-back after write
grep -rn "read_back\|verify_written\|assert .*exists" src/ --include='*.py' | head -n 5

Anti-patterns

BadGood
output.mp4 sent with the run described in chatsteadycut_run3_ride42_hfov120.mp4, parameters also drawn on the first frame
"Wrote 37 facts" as proof of persistenceDump them back from the store and count again
Cache keyed on the request onlyKey includes source identity; invalidated when it changes
"Pushed, so it's done"Read back the merged state, or say "sent, not yet landed"

Related

  • evidence-discipline GC-68 — verify the artifact you deliver is the one you measured.
  • factual-integrity — a claim with no retained source is unverifiable; this skill keeps the source.
  • project-discovery — where the project's own artifact and cache conventions live.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.