ワンクリックでManusで任意のスキルを実行

$pwd:

estimation

Name: Estimation
Author: aj-geddes

// Guides the user through risk-based task estimation using decomposition, three-point estimates, and PERT formula. ALWAYS trigger on "estimate", "how long will this take", "time to complete", "sizing", "scope this", "effort estimate", "how many hours", "story points", "project timeline", "when will this be done", "cost estimate", "level of effort". Use when sizing any work item, feature, or project.

Manusで実行

$ git log --oneline --stat

stars:8

forks:6

updated:2026年3月10日 10:15

ファイルエクスプローラー

5 ファイル

SKILL.md

readonly

name

estimation

description

Guides the user through risk-based task estimation using decomposition, three-point estimates, and PERT formula. ALWAYS trigger on "estimate", "how long will this take", "time to complete", "sizing", "scope this", "effort estimate", "how many hours", "story points", "project timeline", "when will this be done", "cost estimate", "level of effort". Use when sizing any work item, feature, or project.

Estimation

Process

1. Decompose Exhaustively

Break tasks until each piece is < 8 hours and estimatable with confidence.

Task: "Add user authentication"
  ├─ Database schema (2h)
  ├─ API endpoints (16h) → decompose further
  ├─ Password handling (8h)
  ├─ JWT tokens (19h) → decompose further
  ├─ Middleware (14h) → decompose further
  ├─ Testing (23h) → decompose further
  └─ Integration (22h) → decompose further

Rule: If you can't estimate confidently, decompose further.

2. Identify Unknowns

unknowns:
  technical:  ["Never used bcrypt", "Don't know JWT refresh pattern"]
  domain:     ["Password rules unspecified", "SSO needed later?"]
  external:   ["DBA review has 2-day SLA"]
  resource:   ["Frontend dev availability uncertain"]

Every unknown adds risk. Quantify it.

3. Three-Point Estimate

For each atomic task:

Estimate	Meaning
Optimistic (O)	Everything goes perfectly
Realistic (R)	Normal conditions, typical hiccups
Pessimistic (P)	Murphy's Law applies

"Implement password hashing"
O: 1h  (copy existing pattern)
R: 3h  (read docs, test, handle edge cases)
P: 8h  (version issues, need alternatives)

Rule: If P > 3x R, decompose further.

4. Calculate PERT

Expected = (O + 4*R + P) / 6

Example: (1 + 4*3 + 8) / 6 = 3.5h

5. Apply Risk Buffers

Risk Level	Multiplier	When
Low	1.0-1.2x	Well-understood, clear requirements
Medium	1.2-1.5x	Some unknowns, moderate complexity
High	1.5-2.0x	New tech, unclear requirements
Critical	2.0-3.0x	Multiple unknowns, bleeding-edge

Risk categories:

Category	Multiplier	Triggers
Technical	1.3-2.0x	New lang/framework, complex algorithms, perf requirements
Domain	1.4-2.5x	Unclear requirements, regulatory, stakeholder disagreement
External	1.5-3.0x	Third-party deps, vendor timelines, cross-team
Resource	1.2-2.0x	Availability uncertain, skills gap, access issues

6. Add Integration Buffer

Integration is where 50% of bugs live. Always add 20-30%.

Subtasks total: 40h
Integration buffer (25%): 10h
Final: 50h

Anti-Patterns

Anti-Pattern	Problem	Fix
"2 Hours"	"I have no idea"	Use three-point + PERT
Secret Padding	Destroys trust	State explicit buffer: "5h + 5h buffer"
Ignoring Integration	Underestimate 30-50%	Always add integration buffer
Forgetting Testing	"Done" != done	Include test time (often 1:1 with dev)
No Confidence Level	False precision	Use ranges: "45h +/- 5h"

Communication Template

[ESTIMATE] (+/-[UNCERTAINTY]) assuming [ASSUMPTIONS]

Breakdown:
- Component 1: Xh
- Component 2: Yh
- Integration: Zh

Confidence: [High/Medium/Low] (XX%)

Assumptions:
1. [Critical assumption]

Risks:
- [Risk]: [impact] / [mitigation]

Dependencies:
- [External dependency]

Unknowns:
- [Unknown] - [how to resolve]

Validation Checklist

Before finalizing any estimate:

When to Re-Estimate

Requirements change -- re-decompose affected components
Unknown resolved -- update with actual findings
Overrun > 20% -- stop and reassess remaining work
Dependency changes -- adjust timeline
Team changes -- different skills = different estimates

Script

Run the interactive PERT estimation helper:

skills/estimation/scripts/estimate.sh
# or
./scripts/estimate.sh [--output filename]

References

references/decomposition-examples.md -- full worked examples
references/risk-analysis.md -- risk assessment framework
references/communication-guide.md -- stakeholder communication patterns
scripts/estimate.sh -- interactive PERT estimation script

related-skills.json

同じリポジトリ

go.md

from "aj-geddes/unicorn-team"

Guides Go development with idiomatic patterns, tooling, and project structure. ALWAYS trigger on "go project", "go module", "go mod", "goroutine", "channel", "go test", "go build", "golangci-lint", "go interface", "go error handling", "go concurrency", "go struct", "go anti-pattern", "go best practices", "go tooling", "go lint". Use when setting up Go projects, writing idiomatic Go, choosing concurrency patterns, or configuring tooling. Different from testing skill which covers general test strategy; this covers Go-specific testing patterns and tooling configs.

2026-03-248

orchestrator.md

from "aj-geddes/unicorn-team"

Coordinates the 10X Unicorn agent team. ALWAYS trigger on "implement", "build", "create", "design system", "deploy", "learn new language", "refactor", "fix bug", "set up CI", "code review", "how long will this take", "estimate", "architecture", "add feature", "write code", "debug", "review PR", "set up pipeline", "migrate", "optimize". Use for any multi-step task, implementation request, architecture decision, or quality enforcement. Different from individual agent skills which handle execution -- this skill handles coordination, routing, and quality gates between agents.

2026-03-178

domain-devops.md

from "aj-geddes/unicorn-team"

Guides the user through containerization, CI/CD pipelines, Kubernetes deployments, observability, and infrastructure management. ALWAYS trigger on "dockerize", "CI/CD", "kubernetes", "deploy", "monitoring", "logging", "metrics", "helm", "infrastructure", "observability", "rollback", "scaling", "pipeline", "container", "k8s", "GitOps", "Dockerfile", "health check", "troubleshoot deployment". Use when containerizing applications, building pipelines, deploying services, setting up monitoring, or debugging infrastructure issues. Different from the DevOps agent (agents/devops.md) which handles orchestration and runbook execution rather than pattern guidance.

2026-03-158

code-reading.md

from "aj-geddes/unicorn-team"

Strategic code comprehension protocol for navigating existing codebases. ALWAYS trigger on "understand this code", "how does this work", "refactor", "before changing", "legacy code", "code review", "what does this do", "explain this codebase", "walk me through", "unfamiliar code", "read through". Use when exploring unfamiliar code, preparing to modify legacy systems, debugging complex issues, or conducting code reviews. Different from pattern-transfer which maps known solutions across domains -- this skill builds accurate mental models of existing code.

2026-03-108

javascript.md

from "aj-geddes/unicorn-team"

Guides JavaScript and TypeScript development with modern tooling and patterns. ALWAYS trigger on "typescript", "javascript", "tsconfig", "eslint", "prettier", "vitest", "jest", "react", "node.js", "express", "npm", "package.json", "type error", "ts error", "js project setup", "frontend testing", "react hooks", "discriminated union", "type guard". Use when configuring TS/JS tooling, writing tests, structuring React or Node apps, or debugging type issues. Different from testing skill which covers general strategy; this covers JS/TS-specific test patterns and framework configs.

2026-03-108

language-learning.md

from "aj-geddes/unicorn-team"

Guides rapid acquisition of new programming languages, frameworks, and paradigms through a structured 5-phase protocol. ALWAYS trigger on "learn Rust", "learn Go", "new language", "pick up a language", "never used X before", "getting started with", "how does X work", "teach me Y", "unfamiliar technology", "language comparison", "evaluate language", "switch to X", "polyglot", "first time using". Use when encountering unfamiliar technology, evaluating language choices, or onboarding to a new stack. Delegates to Polyglot agent for execution.

2026-03-108

package.json

"author": "aj-geddes"

"repository": "aj-geddes/unicorn-team"

GitHub リポジトリを開く Creator のリポジトリを見る

$ install --global

$ download --local

Manusで実行

$ useful --forSOC

プロジェクト管理専門家ビジネス・金融業務職13-1082L4

estimation

Estimation

Process

1. Decompose Exhaustively

2. Identify Unknowns

3. Three-Point Estimate

4. Calculate PERT

5. Apply Risk Buffers

6. Add Integration Buffer

Anti-Patterns

Communication Template

Validation Checklist

When to Re-Estimate

Script

References

このリポジトリの他の Skills

このリポジトリの他の Skills

Estimation

Process

1. Decompose Exhaustively

2. Identify Unknowns

3. Three-Point Estimate

4. Calculate PERT

5. Apply Risk Buffers

6. Add Integration Buffer

Anti-Patterns

Communication Template

Validation Checklist

When to Re-Estimate

Script

References