replicate
Reconstruct a result as a discovery DAG, freeze it, then have agents independently re-implement from that specification without the original code.
Skill by HudoGriz
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/hudogriz-replicate
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/replicate
What's inside
- Skill name
replicate- Description
- Reconstruct a result as a discovery DAG, freeze it, then have agents independently re-implement from that specification without the original code.
- License
- MIT
- Instructions
- 24 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Discovery DAG and blind replication The object under verification is the path from inputs to claim, not only the final number. ```bash arh dag init -n N arh dag check -n N arh dag freeze -n N ``` The DAG should expose orphan nodes, branch counts and decisions that narrative reports can hide. Once frozen, it becomes the replication specification. ```bash arh replicate -n N --agents 3 --harnesses codex,opencode,claude arh replicate report -n N ``` Each agent runs with a fresh home directory (`replicate_home = fresh` in `.arh/config/harnesses.md`): your harness's memory, history and plugins stay out of a session that is meant to be blind, and only the login files listed in `fresh_home_files` are linked in. If a harness cannot log in that way, add its login file to that list rather than switching to `inherit`; an inherited home is named in the report as unverified blindness. Agents report values, set membership and ambiguities they had to resolve. Treat agreement as evidence rather than confirmation; the useful signal in a fan-out is disagreement, contested membership and specification ambiguity.
From HudoGriz/autoresearch-hpc, MIT.
More from HudoGriz/autoresearch-hpc
- 0.561score
cross-check
HudoGriz
Have a different agent harness review an iteration as adversary, re-implementer, estimand auditor, or gotcha scanner before concluding it.
SkillAFreshness: AClaude Code +9 - 0.561score
iterate
HudoGriz
Run one complete AutoResearch HPC iteration — claim a number, pre-declare the design before any result exists, execute, cross-check with a foreign harness…
SkillAFreshness: AClaude Code +9 - 0.561score
verify
HudoGriz
Re-examine an object an iteration already produced by recalculation, re-implementation, or an alternate analytic choice.
SkillAFreshness: AClaude Code +9 - 0.521score
ledger
HudoGriz
Maintain PROGRESS.md, the single source of truth for an AutoResearch HPC project; render status, check it against disk, and preserve superseded entries.
SkillBFreshness: BClaude Code +9 - 0.521score
arms
HudoGriz
Split one iteration into parallel routes that fail independently, each with its own acceptance criteria, negative control and recorded fate.
SkillBFreshness: BClaude Code +9 - 0.521score
evoke
HudoGriz
Select and invoke an optional external research generator when its specialist workflow materially fits the current iteration, while keeping its code outside…
SkillBFreshness: BClaude Code +9
Signals
- Stars
- 0
- Forks
- 1
- Open issues
- 0
- Last commit (this folder)
- Sep 11, 2026 (1 month ago)
- Last release
- Sep 24, 2026 (1 month ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/HudoGriz/autoresearch-hpc
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from HudoGriz
- 0.561score
cross-check
HudoGriz
Have a different agent harness review an iteration as adversary, re-implementer, estimand auditor, or gotcha scanner before concluding it.
SkillAFreshness: AClaude Code +9 - 0.561score
iterate
HudoGriz
Run one complete AutoResearch HPC iteration — claim a number, pre-declare the design before any result exists, execute, cross-check with a foreign harness…
SkillAFreshness: AClaude Code +9 - 0.561score
verify
HudoGriz
Re-examine an object an iteration already produced by recalculation, re-implementation, or an alternate analytic choice.
SkillAFreshness: AClaude Code +9 - 0.521score
ledger
HudoGriz
Maintain PROGRESS.md, the single source of truth for an AutoResearch HPC project; render status, check it against disk, and preserve superseded entries.
SkillBFreshness: BClaude Code +9 - 0.521score
arms
HudoGriz
Split one iteration into parallel routes that fail independently, each with its own acceptance criteria, negative control and recorded fate.
SkillBFreshness: BClaude Code +9 - 0.521score
evoke
HudoGriz
Select and invoke an optional external research generator when its specialist workflow materially fits the current iteration, while keeping its code outside…
SkillBFreshness: BClaude Code +9
Cite this page
replicate skill. Agent Mart. https://agentmart.dev/skills/hudogriz-replicate (last updated 2026-10-24)