send-notes

send-notes --- Route and send content to Morrison-Lab notes repositories

Persist notes, principles, definitions, examples, derivations, or writing rules to the appropriate Morrison-Lab notes repository (psw, rme, mds, lds, pds, sds, wai, etc.) from any working session or repository.

When this fires

• User says "send this to psw", "psw: ...", "/psw", "add this writing principle to psw"
• User says "send this to rme", "rme: ...", "/rme", "add this regression note to rme"
• User says "send this to mds", "/mds", "send this to lds", "/lds", "send this to pds", "/pds", "send this to sds", "/sds", "send this to wai", "/wai"
• A session generates an insight, teaching example, mathematical derivation, or style rule whose natural home is one of the Morrison-Lab notes or lecture books rather than the current working repo.

Notes Repository Registry

Morrison-Lab maintains focused notes repositories, each with a standardized 3-letter shorthand:

Shorthand	Target Repository	Scope / Topic	Typical Content
psw	Morrison-Lab/psw	Principles of Scientific Writing	Style guide for scientific writing, grammar, conciseness, defining terms, mathematical notation, avoiding clichés, paper organization
rme	Morrison-Lab/rme	Regression Models for Epidemiology	Linear/logistic/Poisson/Cox regression, confounding, effect modification, model selection, survival analysis, epidemiology study designs
mds	Morrison-Lab/mds	Mathematics for Data Science	Linear algebra, multivariate calculus, optimization, matrix decompositions, vector spaces
lds	Morrison-Lab/lds (private)	Machine Learning for Data Science	Supervised and unsupervised learning, deep learning, classification, clustering, dimensionality reduction
pds	Morrison-Lab/pds	Probability for Data Science	Random variables, probability distributions, expectations, conditioning, joint distributions, limit theorems
sds	Morrison-Lab/sds	Statistics for Data Science	Statistical inference, point and interval estimation, hypothesis testing, likelihood theory, resampling, bootstrap
wai	Morrison-Lab/wai	Working with AI	Agentic workflows, prompting, LLM capabilities and failure modes, harness integration, AI-assisted development
win	Morrison-Lab/win	What If Notes	Causal inference, counterfactuals, DAGs, propensity scores, g-methods, target trials
araomm	Morrison-Lab/araomm (private)	Applied Regression Analysis	Lecture notes based on Kleinbaum et al., Applied Regression Analysis and Other Multivariable Methods

(Note: cai / config-ai routes AI configuration and agent policies to ai-config.)

Procedure

1. Identify Target Repository and Topic

Match the content to the appropriate notes repository using the registry above. If the user invoked a specific shorthand (e.g. /psw or /rme), respect that choice. If ambiguous (e.g. a statistical property of an ML estimator could fit sds or lds), choose the repository where the concept is primarily taught or ask the user with a recommended choice.

2. Locate Target Chapter or File

Examine the target repository structure to locate where the content fits:

• Check _quarto.yml, _quarto-website.yml, or _quarto-book.yml to understand the table of contents and chapter organization.
• Inspect the chapters/ directory or root .qmd files.
• Determine whether the content:
  • Fits inside an existing chapter (e.g., adding a new section or callout to chapters/defining-terms.qmd in psw).
  • Merits a new chapter or standalone note (e.g., chapters/<new-topic>.qmd), which also requires updating the navbar/sidebar in the Quarto configuration.

2b. Port existing content by script

If the content already exists in another notes repository, do not retype it. Follow port-by-script: copy the source files with a script, keep ids, wording and citations, and edit only what the target's style rules require. Step 3 applies only to content with nothing to port.

Before porting an ai-config fragment's rule to psw, check that psw does not already state it: run python3 scripts/find-psw-rule-overlap.py <fragment> --psw <psw checkout> from ai-config, which lists the psw files that quote the fragment's example words, and read the files it lists first (grep-is-not-coverage).

3. Draft Content in Target Conventions

Follow the shared Morrison-Lab writing and technical standards:

• Semantic Line Breaks (SemBr): One thought / sentence per line. Never reflow text into arbitrary line wraps.
• LaTeX Math Macros: When writing mathematical notation, use the shared semantic macros from Morrison-Lab/macros (e.g. \E, \Prob, \Var, \Cov, \indic, \R, etc., per use-math-macros).
• Callout Enclosures: Use GitHub or Quarto callouts (> [!NOTE], > [!TIP], > [!IMPORTANT], theorem/lemma environments) where appropriate.
• Tone & Style: Clear, pedagogical, direct, concise, avoiding passive voice and avoiding AI prose tells (per find-ai-tells and psw).
• Pre-review pass: Before the PR goes to a person, run the checklist in notes-pre-review-pass with a separate subagent, and answer every review comment by sweeping its whole class, not the commented line.

4. Deliver on a Dedicated Branch and PR

Never push directly to main of the notes repository. Always isolate work on a dedicated branch and deliver via Pull Request:

Path A: Local Checkout / Worktree Available

If a local clone exists under $HOME/Documents/GitHub/<repo>:

repo_path="$HOME/Documents/GitHub/<repo>"
git -C "$repo_path" fetch origin main
# Create dedicated worktree
wt="$(mktemp -d)"
git -C "$repo_path" worktree add "$wt" -b notes/<slug> origin/main
# Make edits in $wt/...
git -C "$wt" add <files>
git -C "$wt" commit -m "docs(<topic>): <concise description of additions>"
# Push and open PR
git -C "$wt" push -u origin notes/<slug>
gh pr create --repo Morrison-Lab/<repo> --base main --head notes/<slug> \
  --title "docs(<topic>): <concise description>" \
  --body "<summary of content added>"
# Keep worktree intact while driving PR through review; clean up after merge.

Path B: Remote / GitHub API / Temporary Clone

If working remotely or without an existing local checkout:

# Clone shallowly into a temporary directory
tmp_dir="$(mktemp -d)"
git clone --depth 1 https://github.com/Morrison-Lab/<repo>.git "$tmp_dir"
cd "$tmp_dir"
git checkout -b notes/<slug>
# Apply edits to the relevant .qmd file(s)
git add <files>
git commit -m "docs(<topic>): <concise description of additions>"
git push -u origin notes/<slug>
gh pr create --repo Morrison-Lab/<repo> --base main --head notes/<slug> \
  --title "docs(<topic>): <concise description>" \
  --body "<summary of content added>"

5. Drive PR to Clean (ARDI)

• Ensure the PR description follows lab conventions and agent disclosure footer: _Posted by <Agent Name> (AI agent) --- not written by a human._
• Run local formatting / link checks if tooling is present.
• Monitor automated checks and drive review to clean.

6. Wrap Up

• Once the PR is merged into main, remove any temporary worktree or clone: git -C "$repo_path" worktree remove "$wt" (or rm -rf "$tmp_dir").

Relationship to other skills

• push-memory: Routes general-purpose AI agent memories and rules to ai-config.
• config-ai (cai): Routes AI capability and infrastructure requests to ai-config.
• use-math-macros: Standardizes LaTeX macros across all notes repos.
• sembr-reformat: Formats prose to semantic line breaks.
• find-ai-tells: Audits drafts for AI-generated stylistic tells.

Anti-patterns

• ❌ Pushing directly to main on a notes repository without a PR.
• ❌ Dumping unstructured raw text instead of properly formatted Quarto (.qmd) sections.
• ❌ Inventing ad-hoc LaTeX math symbols instead of using shared macros from Morrison-Lab/macros.
• ❌ Modifying unrelated chapters or files in the target notes repository.
• ❌ Forgetting to update _quarto.yml / _quarto-website.yml when adding a brand new chapter file.