Skip to content
Back to skills

Okr Frameworks

ASecurity

Use when writing OKRs, setting quarterly goals, defining key result baselines and targets, or reviewing goal alignment

  • 2 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added October 5, 2026
ai-agentsgotestingdatabase

Security analysis

A100/100

Scanned October 5, 2026

npx -y skills add mariomile/builder-os --skill okr-frameworks --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Okr Frameworks?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Okr Frameworks
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/mariomile-okr-frameworks/badge)](https://www.skillsdirectory.com/skills/mariomile-okr-frameworks)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: okr-frameworks
description: "Use when writing OKRs, setting quarterly goals, defining key result baselines and targets, or reviewing goal alignment"
---

# OKR Frameworks

## Mode and resources

Follow [operating modes and resource paths](../../references/operating-modes.md). **Standalone:** Draft or review the supplied objectives, key results or metric. Where the baseline is unknown, propose measurement before an improvement target; missing company objectives limit alignment claims but do not block a scoped draft, which may state alignment unverified.

**Lifecycle:** Apply the named phase prerequisites, artifact paths and gate recording only when the user requests that phase or initiative. Missing prerequisites block that lifecycle transition, not a standalone artifact. Completion markers with gate verdicts claim lifecycle completion only after the gate passes.

Use `saas-metrics-reference` only for an unclear metric definition, the needed shape in `references/analytics-contract.md` for baseline measurement, and `evidence-ledger` for lifecycle tagging. Retrieve facts before asking; ask what remains per the `pressure-testing` rounds: every question lists the options, recommends one and says why, and a factual question offers ways to close the gap, never guessed values.

Read [the OKR method reference](references/okr-method.md) for objective and KR rules and examples, scoring, the metrics triad, review cadence and all anti-patterns.

## Capabilities

| Capability | Used for | Floor when absent |
|-----------|----------|-------------------|
| `analytics.query` | The baseline behind every key result | Ask the user; tag `[doc:user-{date}-{topic}]`; a KR whose baseline is user-provided says so |
| `db.query` | Database baselines: accounts, revenue, usage | Same floor |
| `docs.search` | Strategy, prior OKRs, the objectives above | Ask for them; without the level above, alignment cannot be checked |
| `docs.write` | Requested document destination or publishing | Use the requested accessible format; publication/config changes require authorization |
| `files.read` / `files.write` | The artifact itself | Always present |

## Procedure

1. **Resolve capabilities and read the level above:** company or group objectives, current strategy, the north star metric, and the previous cycle's OKRs with how they actually scored. Without the level above you are writing goals, not OKRs: say so rather than inventing an alignment.
2. **Derive objectives.** Three at most. Qualitative, time-bound, and written so someone outside the team can tell whether it happened.
3. **Find the baseline for every key result, first.** Baseline before target, always. For each KR, in order: live data (the relevant query shape or database reading, preserving definition, population, window, capture date and source); the vault (metric name plus product name, periodic notes); the user. Where none resolves, record unknown with the measurement method, owner and proposed capture date, and make this cycle's KR "establish the baseline for X by {date}". Never write a KR with a made-up baseline, and never replace an unknown with zero.
4. **Set targets** from the baseline plus the rate the team has actually achieved, adjusted for what is changing. Then apply the KR quality check: `[Verb] [metric] from [baseline] to [target] by [date]`, 2–4 KRs per objective, each scorable under the team's committed/stretch convention.
5. **Check alignment.** Each objective maps to the level above; each key result maps to an owner. Anything that maps to nothing gets cut or gets a reason.
6. **Report** in [the output contract](references/output-contract.md), ending with `## OKR COMPLETE`. Every baseline tagged; every KR whose baseline is user-provided or absent flagged as such on its face.

## Common Mistakes

| Anti-pattern | Example | Fix |
|-------------|---------|-----|
| Task disguised as KR | "Launch new onboarding flow by Q2" | "Increase activation rate from 22% to 35% by Q2" |
| Objective that's a KR | "Grow revenue by 50%" | "Become the go-to analytics tool for Series A PMs" |
| Unknown baseline replaced by zero | "Increase activation from 0 to 40%" without data | Record unknown; establish a sourced baseline before setting the target |
| KRs all measuring same thing | 3 KRs about revenue | Use Metrics Triad: output + input + guardrail |

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.