Index of every page on this site: llms.txt
All skills
hagaiw avatar

/author-blind-evals

@33ac0e9
Run it yourself

Author a small preregistered SkillRoll case pack from a target skill and neutral repository context without seeing prior run outcomes.

  • 4 files
  • 9.1 KB
  • Updated 2 weeks ago
  • GitHub
Use this Skill: https://skilld.dev/gh/hagaiw/skillroll/author-blind-evals

Nothing lands on disk. Nothing to clean up.

Fork this Skill

Edit a local copy. It keeps the author and licence.

SKILL.md

≈39 tokens for metadata: the name and description. ≈380 when used: this file. ≈1.9k more on demand in 3 files.

Description uses 1.9% of example budget

Before choosing a Skill, your Agent reads its name and description. All available Skills share that space.

  • A shorter description leaves more room for other Skills. This entry exceeds our 1% size suggestion.

In our Claude Code example, all Skill names and descriptions share 8,000 characters. This Skill uses ≈155 characters, or 1.9%.

The 1% threshold is a size suggestion. Longer descriptions can still fit.

Your model, settings, and other Skills decide how much text your Agent can read.

Example settings and source

The example uses a 200k-token context and default Claude Code settings. The count includes the name, description, separators, and when_to_use when present. Codex also counts local file paths.

Skit's source and limits: Codex 0.160.1, Claude Code 2.1.292.

Author blind evals

Work from the pinned target, its repository instructions, and the selected skill bundle. Do not read previous SkillRoll reports, suspected failures, maintainer answer keys, or another evaluator's conclusions.

Choose one or two consequential behaviors the skill distinctly owns. Give each case a Markdown .eval.md file with one skillroll metadata fence and Input, World, and Success criteria sections. Keep Input realistic and self-contained, World private and limited to needed external facts, and criteria observable while allowing equivalent good behavior. Replace every generated placeholder and validate the pack offline before freezing its paths and hashes. Hash each finished case once; when the hash tool accepts multiple paths, request them together, otherwise make one request per case. Do not repeat hashes for unchanged files. Do not run inference or rewrite the target skill. Label the frozen handoff Drafted; no live behavior has been observed yet. Explicitly state in that handoff whether the target SKILL.md changed and whether live inference ran.

If starter placeholders were present, identify them in the handoff as a blocking authoring defect and report that the corrected cases replaced them.

Read authoring context for the blindness envelope and preregistration handoff when additional reads are permitted. If the requester supplies a complete envelope and forbids reads, use the instructions above and write only the requested cases.

Source: SKILL.md on GitHub

No rule matched.

skilld matched fixed text patterns in SKILL.md and file names. Patterns miss obfuscated code.

skilld run checks every file with the same patterns. It asks for approval before it loads a Skill with a behavior marked Needs approval.

No third-party reports yet.

Signed by skilld at 33ac0e9. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 6 days ago.

Activeupdated 2 weeks ago

README badge

README badge for hagaiw/skillroll/author-blind-evals