Skip to content

Sensory Skill

Version 1 · Disclaimer

Download the skill · v1 · CC0 · works standalone or behind the router; installing is covered on the Skills page.

The advisory specialist. It answers one question from several angles: what does this ingredient bring to a beer that is still alive, and what will be left of it after months with Brett? It researches native aroma and flavour by form, the compounds that carry them and their thresholds, what fermentation and ageing do to those compounds, how the ingredient has been used in beer and other fermented drinks across cultures and periods (commercial breweries included), and what it pairs with. It designs no preparation and sets no batch doses: every form it recommends names the specialist that would build it, the Maillard Skill or the Treatments Skill. It is called through the router (sensory:<n>) or on its own.

How it weighs evidence

Every claim carries two grades. The lens tier says how close the source situation is to a living, ageing beer; the evidence grade says how good the source is. Relevance never inflates confidence.

Tier Situation Example
L1 House Your own results, from data/house-evidence.md Two beers that differ only in the pairing
L2 Mixed and wild beer Lambic, Flanders, American wild, Brett-aged and farmhouse beers Cherry in lambic
L3 Other drinks that ferment or age with it Clean beer, fruit and country wines, cider, mead, kvass, tepache, historical herb ales Elderberry in a Polish fruit wine
L4 Into alcohol without fermentation Liqueurs, nalewka, nocino, vermouth, bitters, gin botanicals Green walnuts macerated in spirit
L5 Culinary Compound identity and bridge hints only, never a pairing on its own
Excluded Unfermented sweetening or garnish, food pairing, health claims Mango purée in a cocktail

Evidence grades run from A (analytical and sensory studies) through B (technical literature), C (a commercial release with form or timing stated), D (forums) to E (marketing). Commercial precedent is always shown; on grade C-E alone it is labelled inspiration. Findings are reported as the sources state them ("very potent at flameout, does not age"); the calculator's odour-activity values add a ranking beside them.

Modes

Directive Mode Say something like You get
sensory:1 Scope "Green walnuts: any use?" Forms in use across drinks, cultures and periods; which reach a live beer intact; candidate partners; the owner of each form; an analysis brief
sensory:2 Enhance "My gueuze-style blend lacks depth." The gap in sensory terms; three to six ingredient and form candidates ranked by fit, persistence and precedent; a brief
sensory:3 Profile "Profile hibiscus, dried calyces, with pink peppercorn." The ingredient analysis page, catalogue row, registry block
sensory:4 Extend "Add the dried form to the rosehip page." A new version of the same page: new form, partner, section or evidence, or two pages merged
sensory:5 Critique "Will elderflower and blackcurrant clash in this blend?" What dominates, fades, clashes or masks at 0, 6 and 18 months; three tweaks; one spike test
sensory:6 Tasting debrief "Tasted the quince beer at 6 months: …" Its predictions scored; house-evidence lines; corrections to its catalogue and the page

A typical path: sensory:1 to narrow the field, sensory:3 on the brief it returns, then the process specialists build the preparation. Through the router, R2 runs the profile and adds the Maillard and Treatments sections to the same page; R3 runs scope or enhance first so the process specialists work on its shortlist.

The ingredient analysis page

One page per ingredient family: a dried form or a part (rosehip shells, plum stone) is a section of its parent's page, while different plant parts used as different ingredients (elderflower, elderberry) get their own pages, cross-linked. Sections in a fixed order:

  1. Verdict. 2. Key findings, ranked by relevance, then confidence. 3. Forms and their owners.
  2. Sensory profile: native profile, impact compounds with thresholds and their matrix, the beer at 0, 6, 18 and 36 months, the base beer. 5. Precedent by tier. 6. Pairings and influences, ranked, with method and owner. 7. Spike tests. 8. Chemistry (Maillard Skill). 9. Treatments (Treatments Skill). 10. Hazards carried. 11. Sources. Then a changelog.

The page is updated in place (page_version and a changelog row) and published in the Ingredient Analysis section.

The rules that bite

If you want to… The rule Why
Know what pairs with something Ranked by relevance first: your own isolated trials, then repeated wild-beer precedent, then traditions; compound logic alone is a hypothesis A shared compound is a reason to try, not evidence
Trust a threshold Quoted with value, matrix and source; water values flagged Thresholds vary 10-100 times with matrix and panel
Use a commercial beer as precedent Always shown, graded; inspiration unless the producer states form or timing Most brewery copy says "it was used", not "it worked"
Test a pairing Spike tests only: dose a sample, taste on the spot, discard Today's perception; read it through the fate rules for 18 months
Get a dose for the batch Ask the owner (Maillard or Treatments); this skill gives perceptible and dominant levels as evidence Doses and beer-side sums belong to the process specialists
Use a hazardous part Never a part the registry excludes; kernels and stones only via the Maillard A2 route Hazards are carried to their owners, never judged here

Files

File What it is
SKILL.md Modes, research workflow, the analysis-page contract, constraints
references/research-protocol.md Lens tiers, evidence grades, confidence rules, the four research strands, budgets
references/compounds.md Native compounds: fate rules F1-F13 in a mixed fermentation and in ageing; compound table
references/pairing.md Pairing types, clash patterns, ranking
references/sensory.md The catalogue, by registry ID; fills as ingredients are profiled
references/analysis-page.md The analysis page: family rule, frontmatter, sections, versions
references/spike-protocol.md On-the-spot dose-and-taste ladders
scripts/sensory_calc.py Odour-activity values over time, perceptible and dominant doses, spike ladder volumes
data/setup.md Tasting kit and standing context
data/house-evidence.md Your own results with ingredients and combinations (ships empty)

The calculator

python scripts/sensory_calc.py calc.json (or --example --run). For each compound of an ingredient: its content, its level in the beer at each timepoint (from the fate factors), its odour-activity value as a low-high range, and the doses at which it becomes perceptible and dominant. For a spike ladder: grams or tincture volume per sample, the dilution needed below pipette resolution, and the alcohol the tincture adds. It warns on thresholds not measured in beer, missing fate factors and values likely to overpower. It ranks; it does not predict perception.

Versions

Version Date Download Change
1 2026-09-26 ingredient-sensory-1.zip First release: six modes, lens tiers and evidence grades, native-compound fate rules, pairing ranking, the ingredient analysis page, spike tests, sensory_calc.py