All skills
simota avatar

/titan

@f2c9e0a
by shingo imotasimota/agent-skills85 stars
15

Delivering working code fastest via build-first product delivery through minimum agent chains; S/M scopes skip planning and build immediately. Use when compressing time-to-first-PR.

Use this Skill: https://skilld.dev/gh/simota/agent-skills/titan

This session only. Nothing lands on disk.

referencedecision-matrix.md

≈1.3k tokens on demand. Your agent reads this file only when SKILL.md points to it.

Decision Matrix

Purpose: Read this file when Titan must classify a decision, score cumulative risk, choose whether to consult Magi, or record a decision log entry.

Contents

  • Decision classification
  • Decision matrix
  • Risk scoring
  • Risk budget management
  • Magi consultation protocol
  • Decision log format
  • Common decision patterns

Risk-based autonomous decision framework for Titan.


Decision Classification

Classify every decision along two axes:

Impact Level

Level Criteria Examples
Low Affects single file or internal detail Variable naming, code style, test strategy
Medium Affects multiple files or module interface API endpoint design, state management, DB index
High Affects architecture or user experience Framework choice, auth strategy, data model, deployment target

Reversibility

Level Criteria Examples
Reversible Can undo with git revert or simple change Code changes, config updates, test additions
Semi-reversible Can undo but requires significant effort Schema migration, API contract change, dependency upgrade
Irreversible Cannot undo or extremely costly Data deletion, public API breaking change, production deployment

Decision Matrix

Reversible Semi-reversible Irreversible
Low impact Decide immediately Decide immediately Decide + log
Medium impact Decide + log Decide + consult Magi Decide + Magi + risk log
High impact Decide + consult Magi Decide + Magi + risk log Check Risk Budget

Actions: Decide immediately (no log) · Decide + log (add to Decision Log) · Decide + consult Magi (3-perspective analysis) · Decide + Magi + risk log (consultation + risk score) · Check Risk Budget (calculate score, check threshold, proceed or trigger ON_CRITICAL_RISK_BUDGET).


Risk Scoring

Formula

risk_score = scope_of_change × reversibility_factor + external_dependency + security_impact

Component Values

Component Values
scope_of_change (1-3) 1: Single file · 2: Multiple files/module · 3: Architecture/cross-cutting
reversibility_factor (1-3) 1: Fully reversible · 2: Semi-reversible (migration) · 3: Irreversible
external_dependency (0-2) 0: None · 1: Replaceable service · 2: Vendor lock-in
security_impact (0-3) 0: None · 1: Auth logic · 2: Sensitive data (PII) · 3: Security boundary

Score Range

Score Risk Level Action
1-3 Low Proceed freely
4-6 Medium Log decision
7-9 High Magi consultation + log
10-12 Critical Risk Budget check required

Risk Budget Management

All decisions with risk_score ≥ 4 are added to the cumulative risk budget.

Cumulative Score State Action
0-50 NORMAL Proceed normally, log medium+ decisions
51-75 ELEVATED Verbose logging, Magi for all medium+ decisions
76-99 HIGH Magi consultation mandatory for all decisions
100+ CRITICAL Trigger ON_CRITICAL_RISK_BUDGET — present all decisions to user

Budget does NOT reset between phases. Only user approval (via ON_CRITICAL_RISK_BUDGET) resets to 0.


Magi Consultation Protocol

  1. Frame: Present context, options, constraints
  2. 3 perspectives: Logic (technical merit) · Empathy (user impact) · Pragmatism (delivery speed)
  3. Evaluate consensus: All agree → strong confidence · Split → document minority view
  4. Record: Full rationale including minority perspective in Decision Log

Formal MAGI_REQUEST / MAGI_VERDICT usage and Decision Log integration are defined in this file.

Magi Request Format

Decision needed: [What to decide]
Context: [Current phase, constraints, blocking issues]
Options:
  A: [Description] — Pros: [...] Cons: [...]
  B: [Description] — Pros: [...] Cons: [...]
  C: [Description] — Pros: [...] Cons: [...]
Criteria: [What matters most for this decision]

Decision Log Format

## Decision: [Title]
- **ID**: DEC-[NNN]
- **Phase**: [Current phase]
- **Date**: [YYYY-MM-DD]
- **Impact**: Low / Medium / High
- **Reversibility**: Reversible / Semi-reversible / Irreversible
- **Risk Score**: [N]
- **Cumulative Risk**: [Running total]
- **Choice**: [What was decided]
- **Rationale**: [Why this choice]
- **Alternatives Considered**:
  - [Option B]: [Why not chosen]
  - [Option C]: [Why not chosen]
- **Magi Consulted**: Yes/No
  - Logic: [Perspective]
  - Empathy: [Perspective]
  - Pragmatism: [Perspective]
- **Rollback Plan**: [How to undo]
- **Follow-up**: [Any monitoring or validation needed]

Common Decision Patterns

Category Impact Decisions Action
Architecture High Framework, DB, Auth strategy, Deployment target Magi + risk log
Implementation Medium State management, API design, Test strategy, Error handling Decide + log
Tactical Low Variable naming, File organization, Code style, Comments Decide immediately

Source: SKILL.md on GitHub

2 warnings4mo5 checks · Risk SAFE
  • Gen Agent Trust Hub4mo

    The titan skill is a highly structured product delivery orchestrator that uses a phase-based lifecycle to manage complex build tasks. It incorporates several advanced safety and efficiency mechanisms, including a mandatory agent justification gate, a risk-based decision matrix with cumulative scoring, and an anti-stall recovery engine. No malicious patterns, obfuscation, or unauthorized data access were detected.

  • Socket4mo

    1 alert: gptAnomaly

  • Snyk4mo

    Risk: LOW · No issues

  • Runlayer6mo

    2/9 files flagged

  • ZeroLeaks5mo

    Score: 93/100 · 2 sections analyzed

Signed by skilld at f2c9e0a. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 2 days ago.

Activeupdated last month

README badge

README badge for simota/agent-skills/titan