biascheck
Skill1 starsChanged 4 months agoMITView on GitHub
Use when the user wants an independent authenticity check of a piece of writing, a de-slop second opinion, or a read of whether a draft reads as AI-generated.
In this repository10 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- llms.txtllms.txt
- crew-consultSkill
- crew-reviewSkill
- experimentSkill
- multiplechoiceSkill
- priorartSkill
- secure-researchSkill
- unbiasSkill
The file
Files
biascheck
Overview
An independent authenticity check of a draft: several neutral reads by one model (the balanced tier, which the dispatching agent maps to a model and effort) score it 0-100 against a corpus-research reference and the median is reported with the grounded tells. Report-only. Run the one script; do NOT perform the review by hand.
Usage
Require a file: /biascheck <file> [--reads <n>] [--read family[:tune]] [--item <label>] [--timeout <sec>]
Resolve the script
The script lives at <this skill's base directory>/../../scripts/crew.mjs. Resolve it from the base
directory the harness gives you for this skill; never guess or hardcode another location.
Run it
Echo the exact command, then run it from the current repo root:
node <resolved-path>/crew.mjs biascheck <file> [flags…]
With no flags it runs 3 reads on the codex CLI and reports the median
authenticity score. --reads raises the read count for a steadier number on borderline drafts.
After it finishes
- Surface compactly: the anchor, the READS line, the AUTHENTICITY median and spread, and each read's score, reasoning, and tells.
- Report-only: do NOT edit the draft. Applying fixes is the
unbiasskill's job, or the user's. - The score is a signal, not a verdict; a single read is noisy, which is why the median and the spread are both shown. A plain, terse, or non-native-English voice is not itself a tell.
- On exit 1/2, relay the stderr message verbatim and surface any report that printed on stdout.
- The run appends telemetry to
~/.preflight/modelcalls.jsonland writes nothing into the reviewed repo.
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.