artifact-eval
Skill2 starsChanged 3 months agoMITView on GitHub
Use when an event artifact needs a scored judgment rather than a clean-up, or when the user says "score this invite", "is this good enough to send", "evaluate this abstract", "grade the follow-up", "review the event plan". Scores against a type-specific rubric, returns PASS, REVISE, or FAIL with named weaknesses, then revises and re-scores. Judgment and generation stay in separate passes.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- guerrilla2799/ops-and-scale-ostheir own version, first seen Oct 13, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- guerrilla2799/ops-and-scale-os★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository23 other files it gives its agents.
- audience-machineSkill
- conference-activationSkill
- event-and-scale-osSkill
- event-budgetSkill
- event-contentSkill
- event-followupSkill
- event-logisticsSkill
- event-measurementSkill
- event-program-setupSkill
- event-retrospectiveSkill
- event-selectionSkill
- executive-dinnerSkill
- invite-motionSkill
- meeting-engineSkill
- owned-eventSkill
- roundtable-and-salonSkill
- run-of-showSkill
- speaking-and-stageSkill
- sponsorship-negotiationSkill
- staffing-and-briefingSkill
- venue-selectionSkill
- virtual-eventsSkill
- writing-qualitySkill
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 12
Before an invitation goes to a Tier 1 name
- Distinctive: line 13
Before an abstract is submitted
- Distinctive: line 14
Before follow-up sequences are loaded
- Distinctive: line 16
When two drafts need comparing
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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
Artifact Eval
writing-quality catches the tells. This decides whether the thing is good enough to send to a senior executive who will only read one email from you.
When to use
- Before an invitation goes to a Tier 1 name
- Before an abstract is submitted
- Before follow-up sequences are loaded
- On an event plan or sponsorship decision, before it goes to whoever approves it
- When two drafts need comparing
Inputs
- The artifact, its type, and its intended recipient
- Reads: the relevant workspace files, so it is judged against this event rather than a generic standard
Workflow
1. Judge before you revise, in separate passes
The generator never scores its own work in the same pass. Read fresh, score, then revise. Scoring while holding the intent to defend inflates every time.
2. Run the gates first
Any gate failure is an automatic FAIL regardless of score:
| Gate | Fails if |
| Claim integrity | Any number, guest name, or capability that cannot be traced to a source |
| Guest permission | Names a confirmed attendee without permission, or names an unconfirmed one at all |
| Chatham House | Attributes anything said in an off-the-record room |
| Familiarity | Claims a conversation, relationship, or connection that does not exist |
| Writing quality | Any hard fail from writing-quality |
| Compliance | Anything the program's gifting, hospitality, or brand constraints prohibit |
3. Score the dimensions
For an invitation:
| Dimension | Weight | Full marks |
| Why this person | 25 | A clause that could not have been sent to anyone else |
| Sender fit | 15 | Sender seniority matches recipient seniority |
| The room | 20 | Who else will be there, named or precisely characterized |
| The ask | 15 | One clear ask, easy to say yes to |
| Register | 15 | Reads as a personal note. Passes the forward test and the single-recipient test |
| Length | 10 | Three to six sentences for a private event |
For follow-up:
| Dimension | Weight | Full marks |
| Specific reference | 30 | Something they actually said, in their words. This is the whole email |
| Delivery of what was promised | 25 | Attached or linked, not "I'll send it over" |
| Next step | 20 | Specific, with a proposed date |
| Speed | 15 | Written and ready to send inside the tier's window |
| Register | 10 | No inflated warmth, no "the energy in the room" |
For an abstract or session proposal:
| Dimension | Weight | Full marks |
| Non-promotional | 25 | A committee would not read this as a vendor talk |
| Specificity | 25 | A concrete claim or number, not a category overview |
| Speaker credibility | 20 | Evidence this person has done the thing |
| Takeaway | 20 | The audience can say what they will leave with |
| Title | 10 | Promises something concrete. Would draw a room |
For an event plan or decision memo:
| Dimension | Weight | Full marks |
| Arithmetic | 25 | Expected pipeline and cost per meeting shown, with assumptions listed |
| Owners and dates | 25 | Every deliverable has both |
| Measurability | 20 | Meeting target derived, success and repeat criteria stated |
| Risk honesty | 20 | The lean version, the dependencies, and what could go wrong are named |
| Alternative | 10 | A costed alternative exists, especially on a no-go |
4. Verdict
- PASS, 85+ with no gate failures. Send it
- REVISE, 60 to 84. Named weaknesses by dimension, with what would raise each
- FAIL, below 60, or any gate failure. Say whether to rewrite or fix the specific failure
5. Revise surgically
Fix what was named. Do not rewrite what already scored well. Wholesale rewriting on a REVISE loses whatever was working and frequently scores lower.
6. Re-score fresh
Score the revision as though seeing it new. Loop until PASS, or until two consecutive revisions fail to raise the score, which means the problem is upstream. Route it: an invitation that cannot pass on "why this person" usually has a list problem, not a copy problem.
7. Log every cycle
Including the failures. Over a program the log shows which dimension consistently scores lowest, and that is where the real fix belongs.
Output
- The verdict with per-dimension scores and gate results
- On REVISE or FAIL, named weaknesses with what would raise each
- On a revision, the revised artifact and its new score
- A log line in
workspace/measurement/eval-log.md
Rules & quality bar
- Judgment and generation never share a pass
- Gate failures override scores. A beautiful invitation naming an unconfirmed guest is a FAIL
- Score against this event's files, not a generic standard
- Surgical revision
- Two flat revisions means the problem is upstream. Say which skill to re-run
- Log every cycle
Related skills
- Runs after:
writing-quality - Judges output from:
invite-motion,event-followup,speaking-and-stage,meeting-engine,event-selection - Routes failures back to the upstream skill that owns the weakness
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.