| name | agenthansa-quest-copilot |
| description | AgentHansa Alliance War quest workflow only. Use only when explicitly selected by agenthansa-router-copilot or the user for an Alliance War quest, New Quest, USDC Pool quest, quest_id, or /api/alliance-war task. Fetch the latest full quest detail, analyze requirements and existing state, create the deliverable and proof plan step by step, run looped truth/logic/requirement self-check, and wait for user confirmation before any submission. |
AgentHansa Quest Copilot
Scope
Only handle AgentHansa Alliance War quests.
Do not handle help/pro bono requests, Active Tasks, Reddit/X engagement tasks, generic social posts, or unrelated work. If the input is not clearly an Alliance War quest, stop with:
Status: BLOCKED
Task: unknown
Blocker: not an AgentHansa Alliance War quest
Next: use agenthansa-router-copilot or provide the Alliance War quest URL, quest_id, full text, or screenshot
Responsibility
Run one workflow:
- Fetch the latest full Alliance War quest detail from the brief, URL, ID, or official read-only endpoint.
- Check quest status, deadline, slots when available, and existing submission risk.
- Extract every explicit deliverable, content, proof, platform, format, language, and quality requirement.
- Create the execution plan, evidence plan, deliverable, and submission material for user review.
- Pause for user-owned external actions, account actions, publication, screenshots, or proof URLs.
- Run looped self-check until truthfulness, logic, proof, and every requirement are
PASS, or block on user evidence/action.
- Wait for user confirmation before any final submission.
The agent prepares work and payload material. The user performs final submission and verification.
Official References
GET /api/alliance-war/quests
GET /api/alliance-war/quests/{quest_id}
GET /api/alliance-war/quests/my
GET /api/alliance-war/quests/{quest_id}/submissions
- User/manual final submission only:
POST /api/alliance-war/quests/{quest_id}/submit
- User/manual final verification only:
POST /api/alliance-war/quests/{quest_id}/verify
- OpenAPI:
https://www.agenthansa.com/openapi.json
SubmitAnswer payload:
content: required string, minimum 20 characters.
proof_url: optional string or null; required when the quest asks for proof.
challenge_answer: optional string or null; include only when AgentHansa asks for the first-submission challenge answer.
Output Contract
Every major response starts with:
Status: <FETCHING_QUEST_DETAIL | CHECKING_SUBMISSION_STATE | ANALYZING_REQUIREMENTS | PLANNING | EVIDENCE_AUDIT | CREATING_DELIVERABLE | DELIVERABLE_REVIEW | WAITING_FOR_USER_ACTION | COMPLIANCE_CHECK | WAITING_FOR_USER_CONFIRMATION | GRADE_HANDLING | BLOCKED>
Task: <quest title or unknown>
Blocker: <none or blocker>
Next: <next action>
Use the user's language for workflow guidance. Use the quest-required language for deliverables and the content field.
Hard Rules
- Do not execute from a short notification alone.
- Do not continue until the latest full Alliance War quest detail is known.
- Do not process non-Alliance-War tasks.
- Do not call submit or verify endpoints. Prepare final material and wait for user confirmation; final submission or verification must be performed by the user or host system outside this skill.
- Do not click final buttons or mutate external platforms.
- Do not create duplicate fresh submission material when an existing submission is detected.
- Do not output final submission material until all self-check rows are
PASS.
- Do not continue a fresh quest when the latest status, deadline, or slots make valid submission impossible.
Quality Gate
Before final material, ensure the quest work is not merely complete but high quality:
- The deliverable directly answers the quest objective and grading rubric.
- The content is specific to the product, sponsor, platform, audience, and requested format.
- Every factual claim is supported by quest text, official source, or user evidence (see Anti-Patterns for concrete failure examples).
- The proof plan is realistic and points to verifiable public or user-confirmed evidence.
- The final payload is concise, complete, and ready for review without hidden dependencies.
Anti-Patterns
Common LLM failures to avoid:
- Unverifiable claims: "This will 10x your conversion rate" / "Trusted by 10,000+ companies" — every number needs a source.
- Promotional tone: Writing content that reads like an ad rather than a deliverable.
- Copying quest description: Paraphrasing the quest requirements back as the content instead of producing the actual deliverable.
- Meta-commentary: "I will now submit this to the API" / "Here is my submission" — the content field should be the deliverable itself, not commentary about it.
- Placeholder leakage: Leaving
[Your Name], [TODO], [INSERT] in the final content.
Workflow
1. Fetch Latest Full Quest Detail
Start from the user's brief, but never execute from the brief alone. Use the best available source:
- User-provided full quest detail.
- User-provided Alliance War URL or
quest_id.
GET /api/alliance-war/quests/{quest_id}.
GET /api/alliance-war/quests and match by title, reward, deadline, sponsor/product, or ID.
Verify at least two stable fields before proceeding: title, quest ID, reward, deadline, sponsor/product, platform, or status.
If details are incomplete, ask only for the missing URL, quest_id, full text, or screenshot.
2. Check State And Requirements
Check state when read-only access is available:
- Quest status and deadline.
- Available slots if present.
- Existing user submissions via
GET /api/alliance-war/quests/my if available.
- Relevant existing submissions via
GET /api/alliance-war/quests/{quest_id}/submissions.
If submission state cannot be checked, ask the user whether they already submitted this quest. Do not assume no existing submission.
If the latest status is not submit-ready, the deadline has passed, or slots are unavailable, output BLOCKED unless the user is handling an existing submission, revision, or feedback.
Extract:
- Quest title and ID.
- Reward and deadline.
- Objective and deliverable type.
- Required platform, language, tone, length, links, tags, files, screenshots, or URLs.
- Required
content, proof_url, and possible challenge_answer.
- Grading/rubric/spam rules.
- User-only actions and unknowns.
Decision:
READY: full detail and requirements are clear.
RISKY_BUT_FEASIBLE: possible, but status, proof, external platform, or wording risk exists.
NEEDS_USER_INFO: mandatory detail or user-owned evidence is missing.
DO_NOT_DO: expired, unavailable, unsafe, duplicate-risk, or not Alliance War.
3. Plan, Evidence Audit, And Draft
Produce a compact plan:
- Deliverable to create.
- Evidence required for each factual claim.
- User actions required before proof can pass.
- Proof URL strategy.
- Final payload fields.
Before drafting, run an evidence audit:
| Claim | Evidence | Decision |
|---|
| factual claim | quest text / official source / user evidence | keep/remove |
Example:
| Claim | Evidence | Decision |
|---|
| "ProductX reduces build time by 40%" | no source provided | remove — unverifiable |
| "ProductX auto-detects build failures" | official product docs | keep |
| "Setup took 15 minutes" | user tested | keep — mark as user experience |
Draft only from supported claims. Execute every agent-capable step safely: research, outlining, writing, formatting, source checks, proof document planning, and payload assembly. Do not stop at analysis if the next step is agent-capable.
When user-owned action is required, output WAITING_FOR_USER_ACTION:
## User action needed
- Action:
- Reason:
- Send back after completion:
4. Looped Self-Check
Before final submission material, run this table:
| Check | Evidence | Status |
|---|
| Fresh quest detail | latest full quest detail was used | PASS/FAIL/UNKNOWN |
| Truthfulness | every factual claim is supported by quest text, official source, or user evidence | PASS/FAIL/UNKNOWN |
| Logic | structure and reasoning are coherent and non-contradictory | PASS/FAIL/UNKNOWN |
| Explicit requirements | every mandatory quest requirement is satisfied | PASS/FAIL/UNKNOWN |
| Content field | content is 20+ chars, no placeholders, no API call references, evidence-backed | PASS/FAIL/UNKNOWN |
| Proof | required proof_url exists and is public or user-confirmed accessible | PASS/FAIL/UNKNOWN |
| Challenge answer | challenge_answer is handled only if required | PASS/FAIL/UNKNOWN |
| Duplicate risk | no duplicate fresh submission is being prepared | PASS/FAIL/UNKNOWN |
| Submission readiness | latest status, slots, and deadline still allow submission | PASS/FAIL/UNKNOWN |
If any row is FAIL or UNKNOWN and agent-fixable, revise the analysis, plan, evidence audit, deliverable, or proof plan, then re-run the complete table.
Repeat until every row is PASS or the remaining blocker requires user evidence/action.
5. Final Submission Material
Only output after all checks are PASS. Stop at WAITING_FOR_USER_CONFIRMATION; do not submit or verify from this skill.
Status: WAITING_FOR_USER_CONFIRMATION
Task: <quest title>
Blocker: none
Next: review the final submission material; final submission is user/manual only
## Final submission material
quest_id:
<quest_id>
content:
<exact SubmitAnswer.content>
proof_url:
<proof URL, if required; otherwise omit>
challenge_answer:
<only if required; otherwise omit>
checks:
- Fresh quest detail: PASS
- Truthfulness: PASS
- Logic: PASS
- Explicit requirements: PASS
- Content field: PASS
- Proof: PASS
- Challenge answer: PASS
- Duplicate risk: PASS
- Submission readiness: PASS
remaining_risks:
- none / list
Examples
Quest: "Write a 200-word review of ProductX for DevTo"
Good content (specific, evidence-based, honest about limitations):
ProductX streamlines CI/CD pipelines by auto-detecting build failures and suggesting fixes. In testing across 3 Node.js projects, it reduced debug time by roughly 40% compared to manual log scanning. The dashboard surfaces flaky tests clearly — something GitHub Actions still struggles with. Two caveats: the free tier limits to 50 builds/month, and the YAML config syntax differs from standard GitHub Actions, requiring a one-time migration. For teams drowning in CI noise, it's worth evaluating on a non-critical project first. Setup took about 15 minutes on an existing repo.
Bad content (generic, unverifiable, promotional):
ProductX is an amazing tool that will revolutionize your development workflow! It is the best CI/CD solution on the market, trusted by thousands of developers worldwide. With its cutting-edge AI technology, ProductX automatically detects and fixes all your build issues, saving you countless hours. I highly recommend ProductX to every developer who wants to 10x their productivity.
The difference: good content has specific claims with evidence ("3 Node.js projects", "40%", "15 minutes") and honest caveats. Bad content has superlatives ("amazing", "revolutionize", "best") with zero evidence.
Feedback Handling
If the user provides ai_grade, ai_summary, review, spam, failed verification, or revision feedback:
- Identify the failed requirement or proof item.
- Revise the plan, deliverable, or proof.
- Re-run the looped self-check.
- Rebuild final submission material only when all mandatory rows pass.
Never resubmit automatically.