Board of Advisors Skill

A no-ego strategic thought partner: a persona-based board that debates a decision, converges on a recommendation, checks for bias, stress-tests against scenarios, and journals the call with a review date.

Application: Personal Assistant · Database: alh_personal · Namespace: advsr-

Purpose

The advisor skill is the 24/7 sounding board for the loneliness at the top — the strategic or sensitive decisions you can’t yet take to the real board or the team. You seat a small board of persona advisors (mentors you had, mentors you wished for, thought-leaders whose judgment you trust), each with an archetype, a decision style, a standing charter, and a calibrated pushback level. They each write an independent take, then debate, then converge on a recommendation.

The defining principle is calibrated pushback: never sycophancy (an advisor who agrees is dead weight), never devil’s-advocate-for-sport (manufactured opposition that erodes trust). Every debate ends with the board’s converged landing, or a clearly-scoped disagreement the operator must resolve. The skill also surfaces biases — the operator’s, the AI’s, and the blind spots structural to the operator’s seat — and stress-tests the decision against multiple futures rather than the hoped-for one.

A decision journal closes the loop. Every call is recorded with what was decided, why, what’s expected, and a review date; on that date the board revisits what actually happened versus what was predicted, and feeds the lesson back into the personal context system so the board gets smarter over time. As with its sibling skills, the script stores and queries while Claude does the sensemaking — writing each seat’s take in its own voice, running the debate, and surfacing bias; the operator makes the call.

Decision Workflow

Status flows framing → debating → deciding → decided → reviewed; the CLI advances status automatically as artifacts are stored.

Build your board — seat 3–5 persona advisors with add-advisor (archetype, decision-style, pushback level, inspiration, charter). Retire seats that stop earning their place.
Store personal context — capture role, company, ecosystem, competitive stance, priorities, and past decisions with add-context. This is the difference between generic strategy advice and an advisor who has worked with you for years.
Open a decision — open-decision records the question, stakes, operator style, and the operator’s verbatim primer brain dump (stored messy, unstructured).
Interview — grill the operator on assumptions, unknown-unknowns, and what hasn’t been said out loud; store it as an interview note.
Independent takes — each active advisor writes their take with add-take from primer + interview + context only, without seeing the other takes. The first take moves the decision to debating.
Debate — add-debate stages the clash where takes collide, then converges on the board’s recommendation, preserved dissents, and what would change its mind.
Bias check — add-bias-check answers three questions honestly (operator’s biases, AI’s biases, positional blind spots). This moves the decision to deciding.
Scenario stress-tests — add-scenario runs 3–5 “what if” futures across market, competitor, and internal conditions. If the decision only survives the hoped-for future, take that back to the operator.
Decide + journal + review date — the operator makes the call, presented in their recorded style (options-menu, bottom-line, or pushback-then-space). decide stores the outcome, a journal entry, and a review date.
Review on the date — list-decisions --review-due surfaces due reviews; review-decision records what happened versus what was expected and marks it reviewed, then feed the lesson back as a past-decision context doc.

Data Model

Namespace advsr-, extending the alh-* core types in alh_personal. Notes attach to decisions via alh-aboutness.

Type	Purpose
`advsr-advisor`	A persona seat: archetype, decision-style, pushback-level (gentle\|firm\|relentless), inspiration, charter, seat-status
`advsr-board`	A board of advisors — the collection of persona seats
`advsr-decision`	A strategic decision in flight: question, status, stakes, operator-style, outcome, review-date
`advsr-context-doc`	One entry in the personal context system, by kind (role, company, ecosystem, competitive-stance, priorities, past-decision)
`advsr-scenario`	A single “what if” future the decision must survive: condition, impact, likelihood
`advsr-primer-note`	The operator’s messy brain dump captured at open
`advsr-interview-note`	The agent’s interview of the operator
`advsr-take-note`	One advisor’s independent position; owns `advsr-stance` (for\|against\|conditional\|reframe)
`advsr-debate-note`	The challenge-then-converge synthesis
`advsr-bias-note`	Operator’s biases, AI’s biases, positional blind spots
`advsr-journal-note`	Decision journal entries and reviews
`advsr-seat-on-board`	Relation: an advisor holds a seat on a board
`advsr-take-by`	Relation: attributes a take to the advisor who wrote it
`advsr-scenario-for`	Relation: links a scenario stress-test to its decision

Key Commands

Run as uv run python skills/advisor/advisor.py <command>.

Command	What it does
`add-advisor`	Seat a persona on the board (`--archetype --decision-style --pushback --inspiration --charter`)
`list-advisors` / `retire-advisor`	List seats; retire a seat (history preserved, excluded from debates)
`add-context`	Add a personal-context doc (`--kind role\\|company\\|ecosystem\\|...`)
`list-context`	List context docs (`--kind --full`)
`open-decision`	Open a decision and store the primer (`--question --stakes --operator-style --primer`)
`add-note`	Attach a note (interview, primer, debate, bias, journal, general) to any entity
`add-take`	Store one advisor’s independent take (`--decision --advisor --stance`)
`add-debate`	Store the challenge-then-converge synthesis
`add-bias-check`	Store the bias check; advances the decision to `deciding`
`add-scenario`	Store a scenario stress-test (`--condition --impact --likelihood`)
`decide`	Record outcome + journal note + review date (`--outcome --journal --review-date`)
`review-decision`	Close the loop on the review date; marks the decision `reviewed`
`list-decisions`	List decisions (`--status --stakes --review-due --journal`)
`show-decision` / `report-decision`	Full detail; Markdown decision report
`report-board`	Markdown roster + decision stats
`audit`	Run declarative quality checks (flags skipped bias checks, undated decisions, overdue reviews)

How it fits the suite

The board consumes verified findings from the analyst as decision-grade evidence, so debates rest on checked facts rather than the operator’s articulate primer alone. Open decisions surface in the chief-of-staff daily agenda — decisions in framing/debating/deciding that need the operator’s attention, and journal reviews that have come due. Moves generated by the career skill can be taken to the board when they rise to a genuine strategic decision. Reviews feed the personal context system, so the board accumulates judgment over time. See the Personal Assistant application and its siblings: career, analyst, scribe, ops, chief-of-staff, and coach.