Install
$ agentstack add skill-cuteweather-agent-stimuli-exp-design-review ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Stimuli Exp Design Review
Purpose
你只做一件事:审核 /exp_design.md 是否达到可进入 code generation 的标准。
Inputs (Priority)
1) /exp_design.md 2) /paper/ 3) /data/ 4) /repro_check.md(若存在)
Required Outputs
/exp_design_review.md/script/exp_design_review_gate.json
Review Method (Blocking)
必须沿用既有 exp_design 审核思路,至少覆盖以下检查项:
- Evidence Index 完整性:关键事实都能追溯到明确证据定位。
- Parameter Registry 完整性:关键实现参数可落地、无明显缺项、无自相矛盾。
- Fact Ledger 一致性:事实陈述与 paper/data/code 证据一致。
- Phase/condition 语义一致性:phase 职责、条件分支、trial 流程可执行且无错位。
- Data mapping 完整性:数据字段到实现参数的映射可闭环。
- Conflict/Missing 处置:冲突有裁决,缺失有明确阻断说明。
- 外部素材策略一致性:若实验依赖外部视觉素材,策略、追溯要求与 trial 映射完整。
- Grounding 产物一致性:
/paper/grounding/grounding_index.json必须存在于磁盘上(无论有无参考图)。
- 若 paper 中存在可用于 grounding 的参考图,则 expdesign 阶段应已调用精确技能名
stimuli-grounding-segmentation;若产物缺失或全部回退成 textonly,应优先怀疑stimuli-grounding-segmentation没被正确调用或没有被继续重试。 - 若文件中
status = "no_reference_figures"且families为空数组:确认 paper/ 中确实无图片文件,exp_design.md 中所有 family 的effective_reference_basis均为text_only。 - 若 paper/ 中存在图片文件,且
grounding_index.json中所有 family 的effective_reference_basis都是text_only:必须进一步检查这些图片是否真的全是结果图/统计图。只要任一图片中可见 stimulus exemplar、procedure panel、layout scene、response dial、search display 或 memory sample,就必须判为 issue,因为这说明 exp_design 错误跳过了 figure grounding。 - “刺激可由数学参数生成”不能作为通过理由。若 paper 中仍有可见刺激图或流程面板,review 必须要求至少保留 1 条
figure_then_text的 grounding family;并且尽可能产出crop_and_text的 grounding 产物,而不是直接回退到text_only。 - 有参考图就必须产图:若任何 family 的
reference_priority=figure_then_text,则最终产物中必须至少存在 1 张图片产物(mask / overlay / example)。若全部为空,判定为 critical issue,不得以任何理由通过(包括“刺激可由文本参数指定”“无需外部资产”等)。 - 空产物判定流程(当所有有参考图 family 的产物均为空时):
- 检查
retry_attempts是否非空 → 若空,critical issue(未使用--retry-sweep) - 检查
retry_attempts中是否包含phase=fallback_prompt→ 若无,issue(未写入fallback_prompts) - 检查是否存在
reference_role=layout_reference且match_status=matched的兜底记录 → 若无,issue(未做 scene-level 兜底) - 只有当 1-3 全部通过(即已充分重试、已写 fallback prompts、已尝试 scene-level)且仍然无产物时,才可记为 non-critical issue 并允许 codegen 基于文本参数继续
- 严禁将"有参考图但全部分割失败"标记为
no_reference_figures。若grounding_index.json中存在 family 条目(families非空),则grounding_status必须为has_families,不可为no_reference_figures。 - 若文件中包含 family 条目且全部
match_status = "no_match"(即有参考图但全部分割失败):判定为 issue(非自动 pass),要求在 issues 中记录失败原因,并验证 exp_design 阶段已尝试至少一次重试或 prompt 调整。 - 必须检查
retry_attempts字段:若 family 为no_match且retry_attempts为空数组或不存在,说明未使用--retry-sweep,判定为 critical issue,要求重新运行stimuli-grounding-segmentation并带上--retry-sweep。 - 必须检查
fallback_prompts覆盖:若retry_attempts中没有任何phase=fallback_prompt的条目,说明未在grounding_requests.json中写入fallback_prompts,判定为 issue,要求补充fallback_prompts(至少包含 1 条 scene-level prompt)并重跑。 - 必须检查 scene-level 兜底:若所有 object-level family 均
no_match,但没有任何reference_role=layout_reference且match_status=matched的 family 记录,判定为 issue,要求增加 scene-level grounding 并重跑。 - 若文件中包含 family 条目但全部
effective_reference_basis = text_only,同时 paper 图中又存在 stimulus/procedure/layout 画面:判定为 issue,要求重新运行 exp_design / grounding,并补出grounding图。 - 若文件中包含实际 families 条目且任何
effective_reference_basis为crop_and_text:验证paper/grounding/examples/、paper/grounding/overlays/、paper/grounding/masks/中有实际图片文件。若声称crop_and_text但无对应图片,判定为 critical issue。 - 若
grounding_index.json不存在,判定为 critical issue(说明 exp_design 阶段未执行 grounding 扫描)。
Gate Contract (Mandatory)
script/exp_design_review_gate.json 必须包含以下最小结构:
{
"passed": true,
"summary": "string",
"checks": {
"executed": true,
"evidence_index": {"passed": true, "issues": []},
"parameter_registry": {"passed": true, "issues": []},
"fact_ledger": {"passed": true, "issues": []},
"phase_condition_semantics": {"passed": true, "issues": []},
"data_mapping": {"passed": true, "issues": []},
"conflict_missing": {"passed": true, "issues": []},
"asset_strategy": {"required": false, "passed": true, "issues": []},
"grounding_artifacts": {"required": true, "passed": true, "issues": [], "grounding_status": "has_families | no_reference_figures", "retry_sweep_verified": true}
},
"critical_issues": [],
"next_actions": []
}
Hard Rules
critical_issues只要非空,passed必须为false。- 若任一关键检查未执行,
passed必须为false。 exp_design_review.md与 gate 结论必须一致,不得一个通过一个不通过。- 不得用“看起来差不多”替代可核验证据。
Return Contract
按调用方要求返回结构化 JSON。
- 若
exp_design_review_gate.json.passed = true,可返回succeeded。 - 若
passed = false(发现阻断问题),返回failed_retryable,并在error/summary明确 blocker。
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: cuteweather
- Source: cuteweather/agent
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.