Repository navigation
Releases
v2.0.0: Lab Rat
phd-skills is now Lab Rat: new name, rat logo, and repo fcakyon/labrat.
- skills fold the old commands and agents into 11 skills that trigger on their own or run as
/labrat:<skill> - hooks trimmed to a launch checklist reminder, a Stop second opinion and notifications
- launch checks that the code matches a pushed commit before long runs
- install id is now
labrat@labrat
claude plugin marketplace add fcakyon/labrat
claude plugin install labrat@labrat
Coming from phd-skills: run claude plugin uninstall phd-skills@phd-skills and claude plugin marketplace remove phd-skills first.
v1.3.0: Catch AI mistakes before they cost weeks of compute
v1.3.0: Catch AI mistakes before they cost weeks of compute
From paper-writing helper to ML-training discipline plugin in six PRs.
Six of the seven war stories in the README are now actively prevented.
The headliner
A second Claude reviews the first one's conclusions before they reach you.
The Stop hook is now an agent, not a static prompt (#5). Every time the assistant decides it is "done," a fresh-context Sonnet-4.6 spins up in its own session, reads the actual artifacts, and pushes back if the conclusions do not survive contact with reality. 300-second budget, isolated context, zero shared state with the run that produced the work.
Four new skills, zero config
Just describe what you want, they auto-trigger (#7).
| You say | What runs |
|---|---|
| "reproduce this arxiv paper" | Reproduce: 7-stage pipeline from arxiv URL to replication runs |
| "why is my loss diverging" | Debug: evidence-first probes, no guess loop |
| "compare run A to baseline" | Compare: same-epoch alignment across wandb / neptune / tb / mlflow |
| "launch the new training run" | Launch: pre-flight checklist before burning 50 hours of GPU |
Six new silent guardrails
You never invoke these. They catch the assistant being wrong (#7).
- destructive_path_guard: fabricated paths in
rm/mv/dd/ force-push - long_job_launch_inject: pre-flight before multi-hour ML launches
- outbound_artifact_reminder: unverified commands in teammate messages
- remote_inplace_edit_guard: in-place SSH edits to git-tracked files
- jargon_scrub: sprint-codes leaking into public commits
- timezone_scrub: timezone tokens drifting from the system clock
Fixes & polish
- Stop hook infinite loop on bare acknowledgments, fixed by @ccyrene (#8)
- All hook scripts migrated to the
hookSpecificOutputschema (#9) - Star history chart and support acknowledgment in the README (#6)
- README compacted 8% with no loss of UX or USPs (#10)
Install
claude plugin marketplace add fcakyon/phd-skills
claude plugin install phd-skills@phd-skillsZero MCP. Zero config. Works the moment it is installed.
Contributors
Full changelog: v1.1.0...v1.3.0
v1.1.0
phd-skills
Research integrity plugin for Claude Code — paper auditing, citation
verification, experiment analysis, and methodology-first skills for
academic workflows.
Built by Fatih Cagatay Akyon
(1300+ citations, 7 patents) after 200+ Claude Code sessions, tens of
critical AI mistakes caught the hard way, and thousands of hours of
PhD research. Every guardrail in this plugin traces to a real mistake.
Why This Plugin Exists
I use Claude Code daily for my PhD. It's powerful, but it
makes research-specific mistakes that cost hours:
- It typed "done?" as "dont?" and launched an unwanted upload of thousands of files
- It analyzed my full dataset when I asked for a specific 4k/2k/2k split
- It claimed a test covered a bug it had never actually verified
- It never once looked at a figure it generated — just trusted the numbers
Other plugins give you more commands. This plugin gives you guardrails.
Install
claude plugin marketplace add fcakyon/phd-skills
claude plugin install phd-skills@phd-skills
Then run /phd-skills:setup inside Claude Code to configure notifications, LaTeX, and allowlist.
Usage
Open Claude Code in your paper directory, then:
/phd-skills:xray— audit paper against code and data across 5 dimensions, get prioritized fixes/phd-skills:factcheck— verify all BibTeX entries and cited claims against DBLP/phd-skills:fortify CVPR— anticipate reviewer questions, rank ablations, and suggest paper improvements/phd-skills:gaps neural architecture search— find what's missing in the literature/loop 30m check experiment logs, notify me if metrics beat the baseline or if loss starts to diverge"check if my numbers match the code"— skills auto-trigger, no slash command needed"make code publish ready"— prepares code for open-source release with license, docs, and reproducibility checks
After running /phd-skills:setup, all Claude Code notifications (task completion,
background agents) are forwarded to your configured service (ntfy/Slack/email).
What You Get
Commands
| Command | What it does |
|---|---|
/phd-skills:xray |
Audit paper against code and data (5 parallel dimensions) |
/phd-skills:factcheck |
Verify BibTeX entries and cited claims against DBLP |
/phd-skills:gaps <topic> |
Literature gap analysis with web confirmation |
/phd-skills:fortify [venue] |
Select strongest ablations + anticipate reviewer questions |
/phd-skills:setup |
Interactive onboarding (notifications, allowlist, LaTeX) |
/phd-skills:help |
Show all features at a glance |
Skills (auto-trigger — just describe what you need)
| When you say... | Skill activates |
|---|---|
| "design an ablation study" | Experiment Design |
| "find related papers on X" | Literature Research |
| "review my methods section for consistency" | Paper Verification |
| "check if my numbers match the code" | Paper Verification |
| "analyze dataset bias" | Dataset Curation |
| "prepare code for open-source release" | Research Publishing |
| "what will reviewers ask about this?" | Reviewer Defense |
| "setup latex for CVPR" | LaTeX Setup |
Agents (Claude delegates automatically)
| Agent | What it does | Special |
|---|---|---|
paper-auditor |
Cross-checks paper claims vs code and data | Runs in isolated worktree, remembers patterns across sessions |
experiment-analyzer |
Analyzes results from wandb/neptune/local/any format | Can schedule monitoring via cron, sends SSH notifications |
Research Guardrails (run silently — you never invoke these)
| What it catches | Real incident that inspired it |
|---|---|
| Unverified claims, wrong targets, scope creep, dropped requests, assumptions stated as facts | Claude removed introduction novelty claims, analyzed wrong data split, dropped a verification question mid-commit |
| Ambiguous short messages before costly actions | "done?" misread as "dont?", launched unwanted upload |
| Missing citation verification when editing .tex/.bib | Claude propagated unverified author names and venue info |
| LaTeX compilation errors after .tex edits | Errors compounded across multiple edits before being caught |
| Unreviewed generated images/figures | Claude analyzed metrics but never looked at the actual plots |
| Research state loss before context overflow | Long research sessions lost context, leading to rushed conclusions |
How It Compares
| phd-skills | flonat/claude-research | Others | |
|---|---|---|---|
| Commands to learn | 6 | 39 | 13-20 |
| Research integrity hooks | 6 (prompt + command) | 1 | 0 |
| Paper-code consistency audit | 5-dimension parallel | Read-only, no code cross-ref | None |
| Experiment monitoring + SSH notifications | Yes (ntfy/slack/email) | No | No |
| External dependencies | None | npm + pip + MCP servers | MCP required |
| Install time | 30 seconds | 10+ minutes | Varies |
Design Principles
- No MCP dependency — works on any machine, including SSH
- Methodology over scripts — skills teach the approach, Claude generates code for your specific setup (wandb, neptune, local files, whatever)
- Human oversight first — Claude makes premature claims and jumps to conclusions. Every skill builds in verification checkpoints
- Actionable output — ranked suggestions with specific fixes, never just a list of findings
License
MIT — use it, fork it, adapt it to your research.
Built with frustration and care during a PhD at METU.
v1.0.0
phd-skills
Research integrity plugin for Claude Code — paper auditing, citation
verification, experiment analysis, and methodology-first skills for
academic workflows.
Built by Fatih Cagatay Akyon
(1300+ citations, 7 patents) after 200+ Claude Code sessions, tens of
critical AI mistakes caught the hard way, and thousands of hours of
PhD research. Every guardrail in this plugin traces to a real mistake.
Why This Plugin Exists
I use Claude Code daily for my PhD — writing papers, running experiments,
verifying results, managing citations. It's incredibly powerful, but it
makes research-specific mistakes that cost hours:
- It stated my "detached tag head" worked differently than it actually did
- It typed "done?" as "dont?" and launched an unwanted upload of thousands of files
- It silently removed novelty claims from my introduction during a "small edit"
- It analyzed my full dataset when I asked for a specific 4k/2k/2k split
- It claimed a test covered a bug it had never actually verified
- It never once looked at a figure it generated — just trusted the numbers
Other plugins give you more commands. This plugin gives you guardrails.
6 silent hooks catch these exact failure patterns before they cost you time.
8 skills encode research methodology — not scripts. 2 agents build
project knowledge across sessions.
Install
claude plugin install fcakyon/phd-skills
What You Get
Commands
| Command | What it does |
|---|---|
/phd-skills:xray |
Audit paper against code and data (5 parallel dimensions) |
/phd-skills:factcheck |
Verify BibTeX entries and cited claims against DBLP |
/phd-skills:gaps <topic> |
Literature gap analysis with web confirmation |
/phd-skills:fortify [venue] |
Select strongest ablations + anticipate reviewer questions |
/phd-skills:setup |
Interactive onboarding (notifications, allowlist, LaTeX) |
/phd-skills:help |
Show all features at a glance |
Skills (auto-trigger — just describe what you need)
| When you say... | Skill activates |
|---|---|
| "design an ablation study" | Experiment Design |
| "find related papers on X" | Literature Research |
| "write the methods section" | Paper Writing |
| "check if my numbers match the code" | Paper Verification |
| "analyze dataset bias" | Dataset Curation |
| "prepare code for open-source release" | Research Publishing |
| "what will reviewers ask about this?" | Reviewer Defense |
| "setup latex for CVPR" | LaTeX Setup |
Agents (Claude delegates automatically)
| Agent | What it does | Special |
|---|---|---|
paper-auditor |
Cross-checks paper claims vs code and data | Runs in isolated worktree, remembers patterns across sessions |
experiment-analyzer |
Analyzes results from wandb/neptune/local/any format | Can schedule monitoring via cron, sends SSH notifications |
Research Guardrails (run silently — you never invoke these)
| What it catches | Real incident that inspired it |
|---|---|
| Unverified claims, wrong targets, scope creep, dropped requests, assumptions stated as facts | Claude removed introduction novelty claims, analyzed wrong data split, dropped a verification question mid-commit |
| Ambiguous short messages before costly actions | "done?" misread as "dont?", launched unwanted upload |
| Missing citation verification when editing .tex/.bib | Claude propagated unverified author names and venue info |
| LaTeX compilation errors after .tex edits | Errors compounded across multiple edits before being caught |
| Unreviewed generated images/figures | Claude analyzed metrics but never looked at the actual plots |
| Research state loss before context overflow | Long research sessions lost context, leading to rushed conclusions |
Quick Start
/phd-skills:setup # Configure once
/phd-skills:xray # Audit your paper
"design an ablation study" # Skills auto-trigger
"audit my paper against the code" # Spawns paper-auditor agent
How It Compares
| phd-skills | flonat/claude-research | Others | |
|---|---|---|---|
| Commands to learn | 6 | 39 | 13-20 |
| Research integrity hooks | 6 (prompt + command) | 1 | 0 |
| Paper-code consistency audit | 5-dimension parallel | Read-only, no code cross-ref | None |
| Experiment monitoring + SSH notifications | Yes (ntfy/slack/email) | No | No |
| External dependencies | None | npm + pip + MCP servers | MCP required |
| Install time | 30 seconds | 10+ minutes | Varies |
Design Principles
- No MCP dependency — works on any machine, including SSH
- Methodology over scripts — skills teach the approach, Claude generates code for your specific setup (wandb, neptune, local files, whatever)
- Human oversight first — Claude makes premature claims and jumps to conclusions. Every skill builds in verification checkpoints
- Actionable output — ranked suggestions with specific fixes, never just a list of findings
License
MIT — use it, fork it, adapt it to your research.
Built with frustration and care during a PhD at METU.