defossil

Remove prompt-history artifacts from generated text using a line-by-line litmus test.

1|Updated May 28, 2026
One-click install
npx skills add https://github.com/SilvrDuck/skills --skill defossil
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: defossil
Source: https://github.com/SilvrDuck/skills/tree/main/skills/defossil
Command: npx skills add https://github.com/SilvrDuck/skills --skill defossil

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Defossil removes conversation-residue text that only exists because the artifact was produced iteratively, helping your README, docs, comments, and descriptions read like a clean one-shot deliverable.

Core Features & Use Cases

  • Fossil detection by litmus test: Keeps content a stranger would include after receiving the final spec, and flags anything that’s really explained by prior chat history.
  • Targets common fossil types: Transformation labels, scaffolding announcements, defensive reassurance, “now uses X” framing, changelog-in-code sections, TODOs referencing “we discussed”, and similar artifact-narration remnants.
  • Rewrite instead of blunt deletion: When removal would destroy real information, it rewrites fossil framing into product truth (without the “as we changed” provenance).

Quick Start

Use /defossil to scrub an artifact (like a README, docstring, code comment, commit message, or PR description) so it no longer looks like it was iterated on in chat.

Frequently Asked Questions about defossil

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I remove chat history artifacts from generated documentation?

Remove prompt-history artifacts from generated documentation by applying a line-by-line litmus test that classifies text as fossil, borderline, or keep, ensuring deliverables read like a clean one-shot output.

Why does my README contain iterative scaffolding announcements and defensive reassurance text?

Iterative scaffolding announcements and defensive reassurance text appear because the artifact was produced iteratively in chat; these fossil remnants are flagged and rewritten into product truth without the original provenance.

What is the best way to clean up changelog-in-code sections and transformation labels from code comments?

Clean up changelog-in-code sections and transformation labels from code comments by using a taxonomy to detect fossil framing, then rewriting or removing it while preserving genuine invariants and non-obvious behavior explanations.

How do I make commit messages and PR descriptions look like they weren't iterated on?

Make commit messages and PR descriptions look like they weren't iterated on by scrubbing conversation-residue text, removing "now uses X" framing and TODOs referencing prior discussions so they read as clean final specs.

Does removing fossil framing from docstrings risk deleting genuine invariants and behavior explanations?

Removing fossil framing from docstrings does not risk deleting genuine invariants, because the process rewrites fossil framing into product truth instead of blunt deletion, preserving non-obvious behavior explanations.