add-map — Claude Code skill for creating PLSI maps
The add-map Claude Code skill adds a PLSI map for a measured quantity that installed maps do not cover. It activates when a bot reports a map gap or a user asks to track a new quantity.
MIT ★ 93 updated 2026-11-15
npx skills add Insilico-org/plsi --skill add-map -a claude-codeWhat add-map does
The skill checks installed maps for an existing coordinate that measures the same quantity, then creates a user map for anything missing. It defines each coordinate’s unit, code, and inference boundaries; ratios and equations of existing coordinates belong to processor/add-score. The plsi maps validate, plsi maps install, and plsi project commands validate, install, and project the map.
The map is written to <PLSI_HOME>/work/<thread>/user_<topic>.yaml. LOINC codes are looked up with plsi refs show nlm_clinical_tables_loinc, and validation uses --live. PLSI and its plsi command are required.
Age-related values and genetic markers stay local; maps do not send them to the Mothership or an oracle. Shipped maps must not be edited. A user map keeps claim_level: designed until a prediction on it has been scored.
What it helps with
- add a measured quantity missing from installed PLSI maps
- check whether an existing coordinate already covers a quantity
- define units, codes, and inference boundaries for a coordinate
- validate and install a user map, then project it
Example prompts for the agent
- Add a PLSI map for tracking a new quantity and check the installed maps first.
- Create a user map for this measurement with matching units and LOINC codes.
- Validate and install
user_<topic>, then run a PLSI projection.
How to install add-map
1Skills CLI (Vercel)
npx skills add Insilico-org/plsi --skill add-map -a claude-code2Manually: put the files in place
~/.claude/skills/add-map/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/add-map/
1Skills CLI (Vercel)
npx skills add Insilico-org/plsi --skill add-map -a codex2Manually: put the files in place
.agents/skills/add-map/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add Insilico-org/plsi --skill add-map -a cursor2Manually: put the files in place
.cursor/skills/add-map/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install Insilico-org/plsi skills/processor/add-map2Manually: put the files in place
.github/skills/add-map/SKILL.md1Skills CLI (Vercel)
npx skills add Insilico-org/plsi --skill add-map -a gemini-cli2Platform CLI
gemini skills install https://github.com/Insilico-org/plsi --path skills/processor/add-mapManually: put the files in place
.devin/skills/add-map/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about add-map
How do I install add-map in Claude Code?
Run npx skills add Insilico-org/plsi --skill add-map -a claude-code. The skill file will be at ~/.claude/skills/add-map/SKILL.md.
What does add-map do?
It creates and installs a user PLSI map for a measured quantity that installed maps do not cover. It then validates the map and runs a projection.
What does add-map need to run?
PLSI and access to its plsi command are required. Map validation uses --live, and LOINC lookup uses plsi refs show nlm_clinical_tables_loinc.
Source text
SKILL.md · MIT
---
name: add-map
description: Use as PLSI Data Processor when PLSI forwards a map gap - a bot needs to track, predict or score a measured quantity no installed map has a coordinate for (an amount, a questionnaire total, a device metric, a lab test) - or the user asks to track one, so it becomes a coordinate in a validated, installed user map.
---
# PLSI add a map
A map is a contract for reading the log (Tier 0: no code, no formula). A computed score is `processor/add-score`;
anything one formula cannot do (a file format, a fitted model) is PLSI Engineer's (`engineer/extend`).
## When to use
- PLSI forwards a `processor/add-map` request from a bot: what it needs to track, in what unit, and why.
- `processor/intake` recorded coded measurements that no map reads, and the user wants them tracked.
- The user asks to track something PLSI does not track yet.
## Steps
1. **Is it really missing?** Read every map, the user's included, and the coverage:
```bash
plsi maps list --instance <instance>
plsi maps show <map_id> --instance <instance>
plsi maps select --instance <instance>
```
A coordinate that already measures the same quantity (another name, another unit): return it in your `result`
and stop. A ratio, index or equation of existing coordinates: `processor/add-score` instead.
2. **Define each coordinate:** one quantity per coordinate, measured the same way: an `id` (lowercase), a plain
`label` that says what and how (e.g. "reported", "calendar-month mean", "stronger hand"), `kind: measured`, the
canonical UCUM `unit` and `predictable: true` unless it is a list of items (`kind: set` with `collects`).
Related quantities go in one map (`user_<topic>`); at most 6 coordinates.
3. **Codes.** Find LOINC codes for exactly that quantity (`plsi refs show nlm_clinical_tables_loinc` for the search
API; property, timing and method must match the label). Put close but different codes in `reject_codes`
(another hand, another time window, packs instead of cigarettes, another method), so they are never read as this
one. A code a shipped coordinate already reads belongs to that coordinate: use it instead (the install refuses
it). No LOINC code exists: accept `{system: free, code: <map_id>.<coordinate_id>}`, and say in your `result` that
`processor/intake` records its values with exactly that code.
4. **Units, windows, boundaries.** Conversions only with a cited factor (`sources`); otherwise none. `recency_days`
and `score_window_days` from how often the quantity is measured and how fast it changes (shipped maps of the same
kind are the default). `inference_boundaries`: what the map supports (tracking and predictions on the quantity)
and what it does not (diagnosis, comparing methods or devices, values estimated instead of measured or
answered); `min_evidence_class_for_actions: hypothesis_only`; `claim_level: designed`.
**Values that give the age back stay local.** A value in years (a clock or organ age, an age difference) is kept
on this computer by itself; anything else that encodes age (a percentile for age, a pace of aging against age)
gets `local_only: true`, so it never goes to the Mothership or an oracle. A user map can never mark a value
shareable (`map.user_local_only`).
**Genetic markers never leave.** A coordinate named by an rsID or an array probe (`rs429358`, `cg00000029`) stays
on this computer whatever the map says: genotypes and CpG-level values are never shared. Name it by the marker.
5. **Write** `<PLSI_HOME>/work/<thread>/user_<topic>.yaml`, shaped like a shipped map (`plsi maps show
sleep_activity`), with a first comment line saying when the codes were checked. Nothing about the user in it:
no names, dates of their records, values or story; labels and boundaries are generic, so the map can be shared
later as it is.
6. **Validate, install, project:**
```bash
plsi maps validate <PLSI_HOME>/work/<thread>/user_<topic>.yaml --live
plsi maps install --instance <instance> --file <PLSI_HOME>/work/<thread>/user_<topic>.yaml
plsi project --instance <instance> --map user_<topic>
```
A violation is fixed in the file and validated again; `map.code_taken` means the quantity already lives in that
shipped coordinate (step 1).
7. **ONE `result`** to PLSI: the map id, each coordinate with unit and codes, what the projection found (values
present, or the gap and its suggestion), the new `states` hash, and one plain line for the user ("PLSI can now
track your <quantity>"). The requesting bot's work continues with the new coordinate.
## Rules
- Never change a shipped map; a better shipped map is a change to the repository, named in your `result`.
- Never invent a quantity the request does not need, or a code that does not resolve (`--live` checks it).
- Changing an installed user map bumps its `map_version`; old predictions keep scoring against their snapshot.
- The map stays `claim_level: designed` until a prediction on it has been scored.
© Insilico-org · MIT · Insilico-org/plsi