mycelium

ALWAYS invoke this skill when a user wants to perform computational analysis on data in this repository. This skill loads project-specific analysis conventions, dataset registries, and methodology requirements (validation checks, sensitivity sweeps, null hypothesis testing) that are not available from general knowledge — you cannot do the analysis correctly without consulting it first. Trigger for: running any named analysis technique (clustering, PCA, UMAP, differential expression, survival analysis, dose-response, model fitting, time-series, batch correction, statistical tests), exploring or investigating patterns in data, continuing or extending a previous analysis, debugging or troubleshooting analytical results (wrong clusters, unexpected patterns, parameter tuning). The user's intent must be to computationally process or model data — not just read it, not just look at a file. Do NOT trigger for: reading or previewing data files without analysis, writing reports or documents, brainstorming ideas, setting up or initializing repositories, installing conventions, ingesting or importing data files, fixing code bugs unrelated to analytical results, or updating documentation.

ALWAYS invoke this skill when the user wants to respond to, address, or act on Codex review comments on a pull request. Trigger phrases include: "respond to Codex", "address the Codex comment(s)", "fix what Codex flagged", "handle the Codex review", "Codex left comments", "reply to Codex", "@codex", "the bot flagged this", "/mycelium:codex-review". Also trigger proactively when the user pastes a Codex review comment or a link to one and asks you to fix it, or when they are iterating with Codex on a PR and want the round to converge. The defining behavior of this skill — and the reason to use it instead of an ad-hoc fix — is that it does not just patch the single line Codex flagged: it generalizes each comment into an underlying ERROR PATTERN and audits the WHOLE branch for other instances of that same pattern, fixing them all in one pass so later Codex rounds don't surface the same mistake one instance at a time. It auto-detects scope (a specific comment if you point to one, otherwise all open Codex comments on the PR), verifies the fixes against the project's tests, and drafts a reply that summarizes both the targeted fix and the branch-wide audit. When it posts, it appends `@codex review` to re-trigger Codex ONLY if it detects the user already has Codex access on the repo (the Codex bot has previously reviewed or commented on this PR); otherwise it asks first. Posting is outward-facing, so it always drafts the reply, shows it, and posts only after you confirm. Do NOT trigger for: a general analysis-aware code review with no Codex comment involved (/mycelium:review), writing NEW analysis code (/mycelium:analyze), generating reports (/mycelium:report), or open-ended brainstorming (/mycelium:ideas).

ALWAYS use this skill when the user mentions "mycelium", "living repo", "crystallize", or "todo idea". ALWAYS use for requests about making projects self-documenting, tracking knowledge/decisions/learnings, capturing insights for later, transferring knowledge across projects, or exploring unfamiliar data. This skill manages a .living/ directory framework that registers analyses, datasets, and decisions. It handles: project initialization ("set up mycelium", "initialize living repo"), knowledge capture ("crystallize this learning", "record this finding"), future work tracking ("add a todo idea", "track this for later"), knowledge transfer ("transfer knowledge", "cross-pollinate", "sync learnings"), dataset ingestion, convention installation, and progressive disclosure knowledge systems. NOT for: report writing, code refactoring, brainstorming, git init, or generic folder organization.

Invoke this skill for ANY request involving brainstorming, ideation, or generating new research ideas. This is the brainstorming and idea generation skill. Use it when the user wants to brainstorm, generate ideas, get fresh perspectives, hear what experts from different fields would think, find creative angles they're missing, explore new analysis directions, or get unstuck. Trigger phrases include: "brainstorm", "ideas", "fresh perspectives", "what would a [expert] think", "I'm stuck", "creative angles", "wild ideas", "what are we missing", "open to ideas", "before we write it up", "persona brainstorm", "new directions", "where to go next". If the user's request is open-ended and exploratory rather than a specific concrete task, use this skill. Do NOT use for defined tasks like running a specific analysis, fixing code, setting up projects, or ingesting data.

Use this skill when the user wants to add, import, or ingest files or data into the project. Trigger whenever someone mentions: adding data/files to the project, ingesting a dataset, importing new data, bringing in files from any source, having received results or files from collaborators or facilities, needing to place files into the data directory, or setting up/initializing mycelium. Covers all file types (CSV, FASTQ, XLSX, TIFF, JSON, FCS, TSV, images, etc.). This skill copies/moves files into the project structure, generates metadata, updates the data manifest, and documents provenance. Do NOT use for: reading or analyzing existing project data, listing datasets, deleting data, downloading from URLs, or editing code files.