Freshness Judge
v2.0.0LLM通用证据新鲜度判断技能。根据时间窗和证据时间信息，判断每条证据属于current/background/stale/undated/malformed哪一类。在搜索结果标准化之后、需要区分当前证据与背景证据时使用。触发条件：现实问题/新闻/政策/市场分析、需要降低"把旧材料当新材料"风险、时间敏感型任务。
⭐ 0· 65·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description (判断证据新鲜度) matches the instructions and reference docs: the skill only needs normalized evidence and a canonical time frame and does not request unrelated credentials, binaries, or system access.
Instruction Scope
SKILL.md describes only parsing/extracting time info, classifying evidence into five labels, producing a freshness_profile, and optionally emitting compensation_queries for an external orchestrator. It explicitly forbids rewriting evidence, searching/supplementing, or reading unrelated system files. No instructions direct data to unexpected external endpoints or request secrets.
Install Mechanism
There is no install spec and no code files to execute — this is instruction-only documentation, so nothing is downloaded or written to disk by the skill itself.
Credentials
The skill declares no required environment variables, credentials, or config paths. All listed inputs are explicit (normalized_evidence, canonical_time_frame, optional freshness_policy) and proportional to the stated purpose.
Persistence & Privilege
always is false and the skill does not request persistent system privileges or modify other skills' configs. The doc expects an orchestrator to optionally act on compensation_queries — that interaction is outside the skill and not automatic here.
Assessment
This skill is coherent and appears to do only what it claims: parse and label timestamps using the provided canonical_time_frame and optional policy. It does not request secrets or install code. Before installing, confirm two operational details: (1) ensure the upstream 'normalized_evidence' producer (evidence-cleaner) is trusted and does not leak sensitive content into the pipeline, and (2) verify how your orchestrator will treat the emitted compensation_queries (the skill only *generates* suggested queries; if the orchestrator auto-runs them, that could trigger additional network activity and data exposure). Also test edge cases (timezones, relative dates like “本周/昨天”, malformed/future dates) and set freshness_policy defaults (conservatism_level, undated_handling, stale_threshold_months) appropriate to your risk tolerance.Like a lobster shell, security has layers — review code before you run it.
analysisvk974xg53sg19przvft90nmjjhn83ndf7evidencevk974xg53sg19przvft90nmjjhn83ndf7freshnessvk974xg53sg19przvft90nmjjhn83ndf7groundingvk974xg53sg19przvft90nmjjhn83ndf7latestvk974xg53sg19przvft90nmjjhn83ndf7searchvk974xg53sg19przvft90nmjjhn83ndf7signal-intelligencevk974xg53sg19przvft90nmjjhn83ndf7time-analysisvk974xg53sg19przvft90nmjjhn83ndf7
License
MIT-0
Free to use, modify, and redistribute. No attribution required.