<!-- PM-Skills | https://github.com/product-on-purpose/pm-skills | Apache 2.0 -->

OKR Writer

An OKR (Objectives and Key Results) set is a quarterly artifact that translates strategy into measurable outcomes a team commits to drive. OKRs are a focus and learning system, not a project plan, KPI dashboard, performance review device, or roadmap wrapper. Done well, they make priorities explicit, force tradeoffs, enable cross-team alignment, and create visible evidence of progress. Done poorly, they generate roadmap theater, compensation gaming, and false precision.

This skill is a coach, not a template filler. It drafts, reviews, rewrites, and audits OKR sets against the empirical consensus drawn from Doerr (Measure What Matters), Wodtke (Radical Focus), Cagan (SVPG team objectives), Castro (outcome-vs-output), Grove (High Output Management), Torres (continuous discovery), and Gothelf and Seiden (Outcomes Over Output).

Supported Modes

Five entry modes support different engagement levels. Mode is detected from user phrasing; default to Guided when ambiguous. State the detected mode at the start of the response.

• Guided (default, moderate engagement) - brief diagnostic, draft, score against rubric, surface issues, ask user to confirm. Selected by phrasing like "help me write OKRs for X."
• One-Shot (low engagement) - produces a complete OKR set in one pass with all assumptions labeled. Selected by --oneshot flag or phrasing like "just draft OKRs from this context."
• Sustained Coach (high engagement) - iterative loop, one component at a time, re-scored each turn until quality threshold met. Selected by "coach me through OKRs for X."
• Audit Only - user pastes existing OKRs, skill scores and critiques, no new drafts unless user asks. Selected by "review these OKRs."
• Rewrite - convert flawed OKRs, feature lists, or roadmap items into outcome-shaped OKRs. Selected by "fix these OKRs" or "convert this roadmap to OKRs."

When to Use

• Planning OKRs at company, department, product, product-area, team, or initiative scope
• Translating parent OKRs or strategy into team OKRs
• Reviewing a draft OKR set for quality (Audit Only mode)
• Reframing feature, roadmap, or initiative lists into outcome-based OKRs (Rewrite mode)
• Preparing OKRs for stakeholder review
• Identifying whether KRs are measurable and evidence-backed

When NOT to Use

• You only need a dashboard spec - use measure-dashboard-requirements
• You only need event tracking - use measure-instrumentation-spec
• You only need an experiment - use measure-experiment-design
• You only need a hypothesis - use define-hypothesis
• The cycle has ended and you need formal scoring with evidence and learning synthesis - use measure-okr-grader
• The team is purely business-as-usual and needs steady-state KPIs, not stretch outcomes - OKRs are the wrong artifact

Instructions

When asked to write or review OKRs, follow these steps:

1. Detect mode

Read the user's phrasing and classify into Guided, One-Shot, Sustained Coach, Audit Only, or Rewrite. Look for explicit signals (--oneshot, "review these," "fix these," "coach me"). Default to Guided when ambiguous. State the detected mode at the start of the response.

1. Run the empowered-team diagnostic (skip in Audit Only when no new drafting is happening)

Ask briefly:

• Are features, projects, or dates already committed for this cycle?
• Can the team change initiatives mid-cycle if KRs are not moving?
• Who decides what gets built, this team or someone else?

Capture the answer as empowerment_signal: empowered | feature-team | mixed | unknown. This affects output framing in later steps. Do NOT refuse to proceed when feature-team signals are present; instead, plan to add a Disclosure section to the artifact.

1. Determine if OKRs are the right artifact

If the request is really a project plan, KPI dashboard, launch checklist, hypothesis, experiment, or status update, redirect to the appropriate pm-skill or chain. Do not force OKR shape onto non-OKR work.

1. Classify operating context

Capture scope (company | department | product | product-area | team | initiative), cycle (quarter | half | annual | launch window | custom), level, and OKR type (committed | aspirational | learning | operational_health | compliance_or_safety). Default cycle is quarterly when context is missing.

1. Extract or infer strategic intent

Identify the parent objective, strategy pillar, customer problem, or business pressure that motivates this OKR set. If none is supplied, ask once before drafting.

1. Separate outcomes from work

Move features, tasks, projects, launches, hiring counts, and activity counts into Initiatives. The OKR is what changes in the world; Initiatives are bets on how to make that change happen. Apply Castro's litmus test: "if it can go in your backlog, it is not an outcome."

1. Draft or improve the Objective

The Objective is qualitative, specific, directional, and cycle-appropriate. It describes a desired state change, not a project. It connects to strategy. It avoids embedded metrics (numbers belong in KRs). It avoids empty adjectives unless the artifact defines what they mean.

1. Draft or improve Key Results

For each KR include: metric definition, baseline (or recommended-to-measure if missing), target, deadline, evidence source, owner where appropriate, indicator class (leading | lagging | guardrail | health | evidence_generation), and confidence (high | medium | low | unknown). Include a guardrail KR for any optimization that could harm a paired metric (engagement vs quality, growth vs retention, speed vs reliability).

Apply the constraint rules in the next section.

1. Map initiatives as bets

Each initiative names which KR(s) it is expected to move and the assumption underlying that expectation. Initiatives are hypotheses, not commitments. Do not list initiatives as KRs.

1. Run the OKR Quality Audit

Score the draft against the rubric below. Surface issues inline rather than burying them in an appendix. For each risk or fail rating, include a specific recommendation.

1. Apply the empowered-team Disclosure (when needed)

If empowerment_signal == feature-team or mixed, add a Disclosure section: "This OKR set frames pre-committed work as outcome bets. If the metrics do not move when the work ships, that is a learning, not a delivery failure. The team's lever this cycle is to keep shipping; the OKR's lever is to update next-cycle planning." Omit this section entirely when the signal is empowered.

1. Surface open questions

Capture any decisions the user must make that the skill cannot resolve from context. Examples: KR measurement window extending past cycle close, initiative phasing decisions, cohort definition boundaries.

1. Note the source of truth

The artifact is a planning input, not the canonical OKR system. Include a source_of_truth field pointing to the user's actual OKR tracker (company OKR doc, Confluence page, dashboard, dedicated platform, spreadsheet, or wherever the live status lives).

1. Finalize for direct use

Remove all skill instruction commentary from the final artifact. The final output should be reader-facing.

Project Memory Contract

Active only when .claude/pm-skills.local.md exists. With no file, ignore this section entirely and behave exactly as described above.

• Reads: phase, active_initiative, and any prior decision artifacts, so objectives align to what has already been committed rather than restating it.
• Writes: the OKR set as a decision artifact.
• Posture: propose the entry and wait for confirmation before writing, unless

memory_auto_append: true is set, in which case append and echo what was written.

• Write discipline: re-read the file immediately before writing, never from the copy that

produced the proposal. If it changed in between, merge your entry into the current state and re-propose rather than overwriting; add only your own entry and leave every other field and section byte-identical. Nothing enforces this at runtime and the file is gitignored, so a careless whole-file write loses another session's work with no way to recover it.

Memory never supplies a baseline or a target. The refusal to fabricate those stands regardless of what is recorded.

Constraint Rules (MUST / MUST NOT)

These rules are non-negotiable. The skill enforces them in every mode.

• MUST measure outcomes (customer behavior change, business KPI delta, operational health change), not features, tasks, milestones, or activity counts.
• MUST NOT silently fabricate baselines, targets, current values, or benchmarks. Mark missing values explicitly as assumption, placeholder, recommended-to-measure, or not-enough-evidence.
• MUST NOT use OKR scores as individual performance ratings or compensation inputs. If the user requests this, refuse and explain.
• MUST include at least one guardrail or counter-metric KR for any KR that incentivizes growth, speed, or volume.
• MUST treat the 0.6 to 0.7 sweet spot as applying ONLY to aspirational OKRs. Committed, compliance, safety, reliability, and contractual KRs target 1.0.
• MUST default to team-level OKRs. Warn when individual OKRs are requested; explain the sandbagging and false-precision risks.
• MUST NOT become the canonical source of truth. Always include a source_of_truth pointer to the user's actual OKR tracker.
• MUST apply the empowered-team Disclosure when feature-team signals are present. Do NOT refuse the user; adjust the framing.

Quality Audit Rubric

The skill applies this rubric to every OKR set it drafts or reviews. Each criterion gets pass, risk, or fail with a one-line rationale.

• Strategic fit: clear link to strategy, parent OKR, or customer problem
• Objective quality: specific, qualitative, tradeoff-guiding (not a slogan, task, metric bundle, or project name)
• KR outcome quality: measures outcome or behavior change (not tasks, features, or milestones)
• Measurement quality: baseline, target, deadline, evidence source present (or marked placeholder)
• Product influence: team can plausibly influence the outcome
• Focus: 1 to 3 objectives, 2 to 4 KRs each
• Guardrails: quality, reliability, or risk considered for any optimization KR
• Alignment: parent, peer, dependency relationships clear
• Operating rhythm: cycle, check-ins, review points explicit
• Integrity: no compensation coupling, no fabricated data
• Empowered-team Disclosure: included when feature-team signal present, omitted when empowered

Anti-Patterns the Skill Detects

The skill scans for these and either refuses, reframes, or surfaces them with a fail audit rating:

• Feature-delivery KR ("Launch X" instead of "Increase Y from A to B") - reframe into outcome KR; move feature to Initiatives
• Task-count KR ("Complete 10 interviews" without a learning outcome) - reframe or move to evidence-generation type
• Vanity metric KR (metric improves without customer or business value) - flag and propose alternative
• Activity objective (objective describes work, not change) - reframe
• Metric-stuffed objective (objective is just KPIs glued together) - reframe
• Too many OKRs (more than 3 objectives, more than 4 KRs per objective) - force ranking
• Cascading theater (parent KR copied locally without ownership logic) - rewrite as networked alignment
• Roadmap wrapper (OKRs reformat the roadmap) - full Rewrite mode
• Missing baseline (target uninterpretable) - mark recommended-to-measure
• Missing evidence source (no one knows where the score will come from) - mark not-enough-evidence
• Lag-only product OKR (team owns revenue without product outcome) - add a leading product-outcome KR
• No guardrail (optimization may damage quality, trust, retention) - add guardrail KR
• Compensation coupling (people will sandbag or hide learning) - refuse and explain
• Individual OKR default - default to team OKRs; warn if individual OKRs are requested
• Unsupported benchmark (universal target without evidence) - flag and ask for source
• Pre-PMF over-metricization (false quantitative precision when learning is the real objective) - reframe as learning OKR

Output Contract (v1.0.0)

• All required sections present in canonical order: Context, Objective, Key Results, Initiatives as Bets, Guardrails and Health Checks, Alignment Notes, Quality Audit, Open Questions, Suggested Next Step
• Disclosure section is present when empowerment_signal == feature-team | mixed, omitted when empowered
• Every KR includes metric definition, baseline (or marked placeholder), target, deadline, evidence source, indicator class, and confidence
• Initiatives are listed separately from KRs and explicitly tied to which KR(s) they aim to move
• At least one guardrail KR exists for any optimization-style primary KR
• Source-of-truth note is present and points to a non-skill location
• Quality Audit covers all rubric criteria with explicit pass / risk / fail ratings
• Markdown only output. No JSON.
• Foundation classification: no phase: field in frontmatter; uses classification: foundation

Quality Checklist

Before finalizing, verify:

• ] Mode detected and stated at the start of the response
• ] Empowered-team diagnostic run when drafting; signal captured
• ] All required sections present in canonical order
• ] Disclosure section included when feature-team signal present
• ] Every KR has metric, baseline (or placeholder), target, deadline, evidence source, indicator class, confidence
• ] At least one guardrail KR for any optimization primary KR
• ] Source-of-truth note present
• ] No fabricated baselines or targets - missing values explicitly marked
• ] No compensation-coupled framing
• ] Quality Audit applied with explicit pass / risk / fail ratings
• ] Anti-pattern catalog scanned - detected anti-patterns flagged or reframed
• ] OKR type classified (committed | aspirational | learning | operational_health | compliance_or_safety)
• ] Skill instruction commentary removed from final artifact
• ] Markdown only - no JSON output

Examples

See references/EXAMPLE.md for a completed OKR set in the storevine sample thread (Campaigns team, Q3 2026), demonstrating Guided mode on an empowered-team product context with a real cross-team alignment dependency. The companion measure-okr-grader skill handles end-of-cycle scoring; together they cover the full quarterly arc.