Skip to content

coroboros/agent-skills / skills/humanize-en

humanize-en

Verified

Remove AI writing tells from English prose while preserving meaning, structure, code, links and quotations. Use for humanize this, remove AI slop and polish the English on inline text or prose files. Optional -f BRAND-VOICE.md applies brand rules with explicit mechanical and semantic coverage. Inherit authorized rewrite scope; audit/propose remains read-only. Structural README work belongs to write-clear-readme.

evals/evals.json

View on GitHub
{
  "skill_name": "humanize-en",
  "evals": [
    {
      "id": 1,
      "name": "inherited-polish-authorization",
      "prompt": "The user asked: Rewrite the supplied README for concise English, preserving all code and claims. The parent writing skill has drafted the authorized target and invokes humanize-en on that same file. Complete the prose pass.",
      "expected_output": "Read the target, perform the authorized prose edit and applicable checks, preserve code/links/claims and report the diff. No second Apply yes/no prompt is issued. Do not extend the target or publication scope.",
      "files": []
    },
    {
      "id": 2,
      "name": "audit-file-stays-unchanged",
      "prompt": "Audit the supplied README for AI writing tells. Propose improvements, but do not modify it.",
      "expected_output": "Return findings and a proposed rewrite/diff, optionally validating a temporary candidate. The original file is byte-identical afterward. A prose filepath and mechanical checks do not turn an audit into edit authority.",
      "files": []
    },
    {
      "id": 3,
      "name": "contextual-pattern-and-missing-parent",
      "prompt": "Humanize this engineering README using our inherited brand rules. It uses important in a quoted API warning and lists three actual platform limits. The brand parent path does not exist.",
      "expected_output": "Consult references/patterns.md to distinguish literal technical content from rhetorical AI tells; preserve the quote and factual list. Refuse to perform a supposedly brand-compliant rewrite with an unresolved parent. Report the exact failed resolution and missing path, and keep dependent edits pending until the brand input is repaired.",
      "files": []
    },
    {
      "id": 4,
      "name": "sparse-facts-stay-sparse",
      "prompt": "Humanize: Gallery 825 serves as the exhibition space. It boasts over 3,000 square feet. Experts believe its programme may affect attendance; the supplied source names no expert or study.",
      "expected_output": "Preserve Gallery 825, over 3,000 square feet and the uncertain attributed attendance claim. Do not invent a study, year, source, room count or causal result. Simplify phrasing and identify the attribution gap without substituting new evidence.",
      "files": []
    },
    {
      "id": 5,
      "name": "explicit-provenance-json-is-not-resolved-rules",
      "prompt": "Validate the draft 'The launch is scheduled for Monday.' with --rules-json pointing to {\"voice\":{\"name\":\"Synthetic\"},\"merged\":{\"forbidden_lexicon\":[\"launch\"]}}.",
      "expected_output": "Reject the provenance-shaped input with an input error and exact --resolved-json guidance. Do not emit clean or silently ignore the nested rules. Preserve the draft while the brand input is invalid.",
      "files": []
    }
  ]
}
rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 194kUpdated Claude CodeCodexVerifiedInstall