The 22 skills · Platform skill · Everything only
AGENT SKILL
annotations: what it makes your agent do, and what it refuses.
annotations answers the question every recurring report generates, what happened in March, on the chart itself. The discipline is the separation: an annotation explains a number and never changes one, so the narrative layer can never become an unaudited compute path.
The job annotations does
The answer usually exists, in somebody's memory, while the chart keeps the spike unexplained. This skill puts the story where the spike is and keeps the two layers apart. Event markers are data with a date, an author and a scope, rendered on the chart. Team notes are attributed shared state, threaded beside the figure rather than in a separate tool. One direction only: data explains itself in checks, and annotations explain the data in words. There is a sibling mechanism that may move a figure, the structured override, and naming it keeps the line sharp: an outlier flagged out of a total changes the number only through the data layer, attributed and reversible, never through a note.
What annotations refuses
A skill is worth what it stops as much as what it produces. These are the refusals written into annotations, taken from the rules and limits in its own file.
- ✓A note that moves a figure. Corrections go through the data layer, get re-validated and re-tie, or the narrative becomes an unaudited compute path.
- ✓An unattributed marker. Every annotation carries who said it and when, because testimony without an author is a rumour on a chart.
- ✓A note that outlives its context. Markers have a lifecycle, and a stale one is retired rather than left to explain a shape it no longer describes.
- ✓Blurring the layers. The chart shows what happened, the marker records what someone said about it, and the checks stay the only arbiter of what is true.
8 lines from the annotations skill file, verbatim
From skills/annotations/SKILL.md, unedited. Kymira skills are plain text you can
read in full before you install them, which is the point of shipping a method rather than a binary.
Corrections that alter data go through the data layer, get re-validated, and re-tie — that's the pipeline's job. An annotation only *points*: "data-entry error corrected in the 03-14 drop." If an annotation could move a figure, the narrative layer would become an unaudited compute path — the exact thing this architecture exists to prevent. One direction only: data explains itself in checks; annotations explain the data in words. There is a *sibling* mechanism that may move a figure, and naming it keeps
The failure modes annotations addresses
No entry in the catalogue of 51 observed failure modes names annotations as the thing that addresses it. The catalogue records failures we have watched happen, so a skill appears there once a failure it covers has actually been observed and written up. Read the catalogue.
The tier annotations ships in
| Tier | Price | What the tier holds | annotations |
|---|---|---|---|
| Free | $0 | honest-dataviz and the eleven doctrine principles, published in full | Not in this tier |
| Core Pack | $199 | the thirteen craft skills, one-time purchase, lifetime updates | Not in this tier |
| Everything | $349 | all 22 skills, both libraries, the template bank, one-time purchase, lifetime updates | Included |
Both paid tiers are a one-time purchase with lifetime updates, and all sales are final: the files are plain text you keep the moment they arrive, so judge the quality on the free skill and the published doctrine first. The pricing section states exactly what each tier contains.
Installing annotations
A skill is a folder with a SKILL.md file in it. Installing one is copying that folder where
your agent looks for skills, with no build step and no dependencies.
mkdir -p ~/.claude/skills cp -R Kymira/skills/annotations ~/.claude/skills/
Or open the unzipped pack in Claude Code and say install Kymira, which installs every skill in your tier, or double-click the installer for your system. Cursor reads the same skill as a rule file: copy annotations.mdc into .cursor/rules/. Skills are read at startup, so start a new session and ask your agent which skills it has available.
Not for
Annotations are testimony rather than evidence, and attributed, dated testimony is as far as a narrative layer can go. This skill cannot confirm what a note claims, and it is a platform skill, so it ships in Everything only.
Last updated 2026-08-20 · Dylan, founder · one of 22 packaged skills, each one a method written down, not a product tour.