The alignment layer behind Familiar

9 strategic skills for grounded AI work, persistent memory, and operational judgment.

Install all skills: npx skills add open-horizon-labs/skills -g -y

Grounding

Understand the problem before solving it. These skills ensure you're working on the right thing.

Ai

/aim

Clarify the outcome you want—a change in user behavior, not a feature shipped. The aim IS the abstraction. Start here to know what you're solving for.

Use when Starting work, when "why are we doing this?" gets fuzzy, when shipping without impact
Outputs Aim statement with mechanism (how it works) and feedback signal (how we'll know)
Leads to /problem-space, /solution-space
Ps

/problem-space

Map what we're choosing to optimize and what constraints we treat as real. The problem space is the terrain where solutions live—get it wrong and you'll optimize the wrong thing.

Use when After clarifying aim, when constraints feel wrong, when "done" keeps moving
Outputs Problem space map: optimization targets, real constraints, assumed constraints
Leads to /problem-statement, /solution-space
Pt

/problem-statement

Frame the specific problem you're solving. Change the statement, change the solution space. A good problem statement opens possibilities; a bad one closes them prematurely.

Use when After mapping problem space, when stuck in local maxima, when solutions feel forced
Outputs Problem statement with clear scope, success criteria, and non-goals
Leads to /solution-space, /dissent

Execution

Build and ship. Move from understanding to working software in users' hands.

Ss

/solution-space

Explore candidate implementations. Fan out cheaply, score against constraints, prune hard. Four levels of response: Band-aid (patch symptom), Local optimum (optimize current framing), Reframe (question the problem), Redesign (eliminate the problem).

Use when Multiple approaches possible, after problem statement is clear
Outputs Ranked solutions with level (band-aid/optimize/reframe/redesign) and rationale
Leads to /execute, /dissent
Ex

/execute

Do the work. The terraform loop: Pre-flight (check orientation), Execute (build), Detect drift (watch for scope creep), Salvage (if drifting), Update terrain (encode learning). Steer or Abort—never thrash.

Use when Solution chosen, ready to build
Outputs Working implementation with documented decisions
Leads to /review, /ship, /salvage (if drift detected)
Sh

/ship

Deliver to users. Optimize the delivery path—the path from merged code to working install. When execution is cheap, delivery is the work. Reduce delivery-path tax: review-to-merge time, merge-to-prod time, manual approvals.

Use when Work complete, ready for users, or delivery feels slow
Outputs Shipped value, delivery path analysis, bottleneck recommendations
Leads to /review (post-ship), /problem-space (next iteration)

Reflection

Capture learning that compounds. Challenge assumptions. Extract wisdom before restarting.

Rv

/review

Check work and detect drift. Invokes metacognitive review to catch scope creep, over-engineering, incomplete implementations, and strategic misalignment. A second opinion before committing.

Use when Before commits, before PRs, when uncertain about approach, after shipping
Outputs Assessment with specific concerns, drift indicators, recommendations
Leads to /salvage (if drift), /execute (if aligned), /problem-space (next cycle)
Ds

/dissent

Structured disagreement mode. Deliberately seek contrary evidence and alternative framings. Useful for avoiding confirmation bias and groupthink on important decisions. "What would make this wrong?"

Use when Making irreversible decisions, when everyone agrees too quickly, before major commits
Outputs Counter-arguments, risks identified, alternative framings, steelman of alternatives
Combines with /problem-statement, /solution-space, /review
Sv

/salvage

Extract learning before restarting. Code is a draft; learning is the asset. Captures what changed your model, what guardrails should exist, what context was missing. Then restart clean.

Use when Work is drifting, approach reversed 3+ times, scope expanding while "done" fuzzing
Outputs Extracted learnings, new guardrails, updated context for fresh start
Leads to /problem-space (fresh start with new understanding)