72 lines
3.4 KiB
YAML
72 lines
3.4 KiB
YAML
name: adjudication_worker
|
|
description: "Worker-executed adjudication task. Reads the staged deliverable from Tool.Gitea, scores it, writes RAG learnings, and submits the adjudication result."
|
|
debug: false
|
|
system: agent_prompt
|
|
|
|
sections:
|
|
- scene
|
|
- agent
|
|
- project
|
|
- history
|
|
- deliverables
|
|
- message
|
|
- instructions
|
|
|
|
steps:
|
|
# -- Step 0: Think -- evaluate the staged deliverable ---------------------------
|
|
- type: think
|
|
max_tokens: 8000
|
|
scene: |
|
|
You are {agent.name}, an independent performance reviewer and constitutional auditor.
|
|
You evaluate completed work against acceptance goals AND the company's corporate charter.
|
|
system: |
|
|
TASK: {task.name}
|
|
|
|
SCOPE -- SINGLE DELIVERABLE REVIEW:
|
|
You are evaluating ONE specific deliverable file located at `deliverable_staging_path`.
|
|
Do NOT evaluate project-wide completion, chapter count, or how many items remain.
|
|
The project may be a multi-part series -- that is irrelevant.
|
|
You are reviewing ONLY the file at the staging path. Judge it on its own merits.
|
|
|
|
Review this deliverable against the acceptance criteria listed below (if any) AND
|
|
the CORPORATE CHARTER (if present in the message).
|
|
|
|
1. CHARTER COMPLIANCE (when a CORPORATE CHARTER is present):
|
|
Cross-reference the deliverable against the charter's authorized domain,
|
|
forbidden activities, and non-goals. Charter violations are grounds for
|
|
automatic failure (score < 80).
|
|
|
|
2. CORPORATE POLICY EXTRACTION:
|
|
Identify ONE formal Standard Operating Procedure or Corporate Policy
|
|
established by this task (if any).
|
|
|
|
3. PROJECT LEARNING:
|
|
Identify ONE specific architectural fact the project established (if any).
|
|
|
|
4. SCORING:
|
|
Grade the worker from 0 to 100 based on the acceptance criteria and charter compliance.
|
|
Use the criteria rubric if provided; otherwise apply craft and quality standards.
|
|
hint: |
|
|
Do not fabricate policies. If it was a routine task with nothing new, the policy is "NONE".
|
|
Write policies in active, authoritative language. No personal pronouns.
|
|
Format: "Standard Practice: ..." or "Policy: ..." or "NONE".
|
|
Score ruthlessly on a 0-100 scale: 80+ means acceptable work that met requirements,
|
|
90+ means good work with quality above expectations, 95+ means exceptional work.
|
|
Below 80 means the work failed to meet core requirements and must be retried.
|
|
IMPORTANT: Score based solely on THIS deliverable's quality, not on project completion status.
|
|
IMPORTANT: Each criterion in the rubric must be scored 0-100 independently.
|
|
|
|
# -- Step 1: Package -- emit adjudication result --------------------------------
|
|
- type: package
|
|
packet_type: AdjudicationResult
|
|
schema:
|
|
corporate_policy: "string or 'NONE' -- formal Standard Practice or Policy established. Active, authoritative language. No personal pronouns."
|
|
project_learning: "string or 'NONE' -- one bullet: factual project architecture or decision established here."
|
|
score: "integer 0-100 -- 0=Complete failure, 80=Acceptable, 90=Good, 95+=Exceptional. Tasks scoring below 80 will be retried."
|
|
score_breakdown: "object -- per-criterion scores as {criterion_name: integer 0-100}. Omit if no rubric was provided."
|
|
justification: "string -- one sentence explaining the score. Reference specific evidence from the deliverable. Cite charter violations if any."
|
|
|
|
adjudication:
|
|
enabled: false
|
|
|