All authors
Claude Skills by emaballarin
github.com/emaballarin27 skills0 installs9 views
- ReportWrite the final summary of an autoresearch run to `./.ar/final_report.md` — trajectory, best commit and its diff, what worked, what failed, and what to try next. Use for `/ar:report`, "write up the autoresearch results", or "summarise the experiment loop". Reads `./.ar/ar.jsonl` from disk, so it works in a fresh session. Writes only the report; never iterates, commits, or reverts, and leaves the loop able to continue.Votes: 0GitHub stars: 4
- ResumeRun the autoresearch experiment loop — propose one change, measure it, keep it only if it beats the noise floor, repeat. Use to resume or continue an autoresearch run, for `/ar:resume`, for "keep iterating", "next experiment", "continue optimising", or whenever `./.ar/ar.jsonl` exists and the loop should advance. Reconstructs everything from disk, so it survives compaction, `/clear`, and a fresh session. Requires `/ar:start` to have run first.Votes: 0GitHub stars: 4
- StartOpen a new autoresearch experiment loop — dedicated `ar/…` git branch, `./.ar/` state, and a measured baseline noise floor so later gains can be told apart from run-to-run scatter. Invoke ONLY when explicitly asked to start an autoresearch run, open an experiment loop, or `/ar:start` — e.g. "autoresearch this", "start an ar run", "set up a loop to minimise X". Do NOT auto-fire on ordinary optimisation work. "Make this faster", "improve the accuracy", "tune these hyperparameters", "benchmark t...Votes: 0GitHub stars: 4
- StatusPrint the current autoresearch run's state — goal, baseline, noise floor, best-so-far, run tallies, budget, branch, and the last few iterations. Use for `/ar:status`, "how is the loop doing", "what's the best so far", or any read-only check on an autoresearch experiment. Reconstructs everything from `./.ar/ar.jsonl` on disk, so it works in a fresh session with no prior history. Strictly read-only — never iterates, edits, commits, or reverts.Votes: 0GitHub stars: 4
- StopEnd the current autoresearch run — append a `status:stopped` sentinel to `./.ar/ar.jsonl`, which is also what halts the external `ar-loop.sh` driver. Use for `/ar:stop`, "stop the loop", "halt autoresearch", or "that's enough iterations". Writes the sentinel and the final report, then prints (does not run) the command to return to the original branch. Never reverts or discards work — every kept commit stays on the experiment branch.Votes: 0GitHub stars: 4
- Figure ComposerCompose one publication-grade multi-panel figure. Entry from a one-line claim + data refs, OR from an existing figure via `derive_outline_task(png)`. Runs a per-figure loop: outline (12-col grid, per-panel ask + label_budget) → fan-out one subagent per panel (each loads `figure-style`) → tile + stamp letters → adversarial composite review with two-tier feedback (Tier-1 outline_revisions / Tier-2 per-panel violations) → regen affected panels, ≤3 rounds. Kernel exposes panel_task / compose_figu...Votes: 0GitHub stars: 4
- Figure StylePublication-grade figure correctness and legibility rules for final-deliverable figures — not every plot. Quick look or iterating on the analysis (EDA scatters, sanity-check histograms)? Plot plainly without this skill. Producing a figure that ships — report, paper, export, or a file you keep — load this skill first and call `apply_figure_style()` — sets a role-mapped font-size ladder, outward ticks, frameless legends, and 300-dpi output. The skill is a checklist, not a house look: data fidel...Votes: 0GitHub stars: 4
- Literature ReviewFind, verify, and synthesize STEM literature across every field — from "what's the seminal paper for X" through full multi-source reviews. Treats journal articles, arXiv/preprints, and conference proceedings as first-class; grounds every claim in a retrieved source, never fabricates DOIs, checks for superseded/withdrawn/refuted work, calibrates confidence to evidence strength (including SOTA/reproduction/ablation caveats for CS and ML), and emits a tidy DOI-pinned BibTeX reference list. Pairs...Votes: 0GitHub stars: 4
- Paper NarrativeJudge and reshape the STORY a paper's figures tell. Input is the work itself — manuscript (or abstract) + figure deck — no hand-written brief. `derive_paper_brief_task(abstract, captions)` builds the prompt whose JSON is pitch/vision/per-figure-claims; a handling-editor reviewer on the full deck returns hook_verdict (would Fig 1 make me send this for review?), arc (hook→mechanism→evidence→application), figure_moves (panels in the wrong figure), missing_panels (concrete analyses to RUN), kill_...Votes: 0GitHub stars: 4
- Paper ReviewHuman-driven peer review of a scientific manuscript with Claude as the second reader. Claude reads the whole paper at once, tests the reviewer's impression against the text, builds a line-anchored ledger of claims and gaps, calibrates the score against the rubric's boundaries, and constrains and error-checks the summary, title and review the reviewer writes. For conference, workshop and journal reviews once the PDF and the review form are in hand.Votes: 0GitHub stars: 4
- Pdf ExploreUse this skill when the user has attached or pointed to a PDF, paper, report, or other document and the answer needs content from more than one place in it: summarize the methods or any other section, compare sections, find where a topic is discussed, read a value or label off a figure or chart, pull tables out as CSV, or find/list/extract every instance of something across the whole document (datasets, benchmarks, citations, figures, table rows, accession numbers — including appendices). It ...Votes: 0GitHub stars: 4
- AuthorReference for writing documents an agent reads — skills, AGENTS.md / CLAUDE.md, and the files they point at. Use when authoring or editing a SKILL.md, writing or pruning an AGENTS.md / CLAUDE.md, deciding whether a skill should be model- or user-invoked, choosing what to inline versus push behind a pointer, or diagnosing a document the agent follows inconsistently between runs.Votes: 0GitHub stars: 4
- DumpConsolidate the current session's progress into the project's auto-memory directory — the "log everything, update state, prepare for resumption" drill. Use when the user says "log everything", "checkpoint", "save progress", "update memory", "dump state", or autonomously when a long session is approaching context saturation, a natural checkpoint has been reached, or substantial non-derivable state has accumulated that would be lost on session end. Works on any project primed with /mf:prime; wr...Votes: 0GitHub stars: 4
- PrimePrime the current project for the mindfunnel workflow — stamp a project-scoped `AGENTS.md` from the bundled stub (if absent), create a project-local `CLAUDE.md` symlink to `./AGENTS.md`, touch an empty `PROJECT.md` if absent, clean up legacy `SOUL.md` / `CLAUDE.md` / `AGENTS.md` symlinks left behind by older primes, and strip legacy `CLAUDE.md`/`AGENTS.md` entries from `.gitignore` so the committed files track cleanly. Run from the project root. Idempotent with a safety guard — leaves pre-exi...Votes: 0GitHub stars: 4
- SetupBootstrap — seed ~/.mindfunnel/ with AGENTS.md, SOUL.md, USER.md, and PROJECT.md.example from the plugin's bundled templates. Use the first time you run the mindfunnel plugin on a new machine, when ~/.mindfunnel/ is missing, or after an mf upgrade that adds links. Idempotent — detects existing files and never overwrites. Also creates a CLAUDE.md symlink pointing at AGENTS.md inside ~/.mindfunnel/, and — for each agent installed — ~/.claude/{CLAUDE,SOUL,USER}.md + ~/.codex/{AGENTS,SOUL,USER}.m...Votes: 0GitHub stars: 4
- SpinupRead project memory and produce a tight "where we are + next action" brief, then wait for direction. Invoke ONLY when the user explicitly asks to get oriented on prior project work (spin up, catch up, get up to speed, where were we) — not to continue an autoresearch loop, which is /ar:resume. Do NOT auto-fire on general project questions, code edits, or unrelated asks. Works on any project primed with /mf:prime; reads from the Claude Code auto-memory dir under ~/.claude/projects/<slug>/memory/.Votes: 0GitHub stars: 4
- AnalyzeExtract insight from a finished study and return a variance-aware adopt verdict — search-space boundaries, sampling density, infeasible fraction, training-curve pathologies, isolation plots, then adopt or not. Use for `/tml:analyze`, "here are my sweep results", "which config should I pick", "did this change actually help", "why do my runs disagree", "read this wandb export", "is this result real", or after `/tml:round`. Works standalone on results it did not produce — a CSV, a JSONL, or a tr...Votes: 0GitHub stars: 4
- AuditRead a training/evaluation pipeline and emit ranked, evidence-graded, priced findings across the algorithmic (A), systems (B), protocol (C) and architecture (D) tiers. Use for `/tml:audit`, "why is my training slow", "how do I speed this up without losing accuracy", "review my training loop", "what am I leaving on the table", "is my pipeline input-bound", "is my eval setup sound", or before committing to a round of optimisation. Modality-agnostic — vision, text, audio, tabular, graph, time-se...Votes: 0GitHub stars: 4
- PlanChoose an operating point on the speed↔quality frontier, fix the training-step budget, and turn findings into an ordered, gated change plan — with a quality floor, named non-negotiables, revert triggers, a measurement contract, and a stop rule. Use for `/tml:plan`, "which of these should I actually do", "I have N GPU-hours and need quality Q", "how long should I train for", "how much accuracy can I trade for speed", "what order should I make these changes in", or after `/tml:audit`. Writes ./...Votes: 0GitHub stars: 4
- ReviewAdversarially review a proposed or applied change to a training/evaluation pipeline: did the claimed mechanism actually engage, is the measurement still valid, what did it silently break. Use for `/tml:review`, "I made training faster, check it", "review this diff before I merge", "why did throughput improve but accuracy drop", "did this speedup actually work", "check my optimiser swap", or on any diff touching a training loop, data pipeline, optimiser, precision, architecture or evaluation p...Votes: 0GitHub stars: 4
- RoundDesign the next round of experiments — scope one goal, classify every hyperparameter as scientific, nuisance or fixed, build the studies, choose search spaces and a sampler, and allocate the trial budget. Use for `/tml:round`, "what experiment should I run next", "how do I test whether X helps", "design a sweep for this", "is this comparison fair", "set up a hyperparameter search", or after `/tml:plan`. Emits a self-contained study bundle under ./.tml/rounds/NNN/ — config matrix, launch stub ...Votes: 0GitHub stars: 4
- GrillA relentless, round-based interview that stress-tests a plan, design or decision against its own design tree until nothing is left silently assumed.Votes: 0GitHub stars: 4
- DeslopA behaviour-preserving cleanup pass over the change under review — strips comment slop, defensive-check slop, type laundering, unrequested fallbacks and style drift from the diff before /code-review. Never repo-wide, never functional.Votes: 0GitHub stars: 4
- Test AuditAudit and consolidate an existing test suite — low-value, implementation-coupled or duplicated tests, duplicated setup, and the test-only production seams they keep alive. Evidence first, then one approved batch of edits. Campaign mode prunes a whole subsystem's test surface.Votes: 0GitHub stars: 4
- Test GateGate for every test being written or changed — four questions it must answer, the junk-pattern check, red-before-green for regression tests, and extend-don't-restructure for the tests already there. Use when adding tests for new code or an experimental condition, writing a regression test for a bug fix, or changing an existing test's assertions. Runs silently; speaks only for a test that fails the gate or for concerns about the surrounding tests, which it reports and notes rather than fixes. ...Votes: 0GitHub stars: 4
- Bib AuditAudit an existing BibTeX bibliography end to end — verify every entry against Crossref, arXiv (via DataCite), DBLP, Semantic Scholar, OpenAlex, the NeurIPS/PMLR/JMLR proceedings indices, the ICLR/NeurIPS/ICML paper lists and OpenReview, correct wrong or fabricated metadata, replace arXiv preprints with their published version of record, merge duplicates, strip non-citation fields, rekey to `<Surname><Year><FirstContentWord>`, tidy and validate — then research only the entries no source settle...Votes: 0GitHub stars: 4
- Review GateFresh-context review of a change against its stated intent, before the change is acted on. Use before a commit (not the per-iteration commits of an autonomous loop the operator started, such as `/ar:resume`), before a release, or before launching a run whose result will be acted on — whenever the change is more than a one-line tweak, and always for an audit script, a release, or a pipeline feeding an experiment. Hands the diff and a written intent, never the implementer's reasoning, to a fres...Votes: 0GitHub stars: 4