/cost-estimate

Estimate the token cost for running an Agent Triforce workflow. Reads the routing configuration from templates/agent-routing.json to determine which model is assigned to each skill and calculates estimated costs.

Step 1: Read routing configuration

Read templates/agent-routing.json from the project. If the file does not exist, use the default routing assumptions:

• code-health: haiku
• security-audit: sonnet
• feature-spec: sonnet
• implement-feature: user's default model (assume sonnet for estimation)
• release-check: haiku
• review-findings: user's default model (assume sonnet for estimation)
• generate-tests: sonnet

Step 2: Determine workflow type

Ask the user (or infer from context) which workflow they want to estimate. Common workflows:

Workflow	Skills Involved
Full feature (PM -> Dev -> QA)	feature-spec + implement-feature + security-audit
Code health scan	code-health
Security audit only	security-audit
Release check	release-check
Implement + review	implement-feature + review-findings
Test generation	generate-tests

Step 3: Calculate estimate

For each skill in the workflow:

1. Look up the assigned model from routing config
2. Look up average token counts from averageTokensPerSkill in the config
3. Calculate: (input_tokens / 1M * inputPer1M) + (output_tokens / 1M * outputPer1M)
4. Sum all skills for total estimated cost

Step 4: Present results

Display a table showing:

Workflow: [workflow type]

| Skill | Model | Est. Input | Est. Output | Est. Cost |
|-------|-------|-----------|------------|-----------|
| ...   | ...   | ...       | ...        | $X.XX     |

Total estimated cost: $X.XX - $X.XX (+/- 30% margin)

Important notes:

• Cost estimates use average token counts from benchmark runs, not real-time prediction
• Actual usage varies based on codebase size, spec complexity, and conversation length
• Estimates have a +/- 30% margin of error
• If a configured model is not available in the user's plan, the system falls back to the most capable available model
• Model override with --model opus applies to a single run only

Configuration file: templates/agent-routing.json