Signal Intelligence Pack
v2.1.0LLM通用前置grounding技能组。在正式分析、判断、报告或规划之前，将问题拆解为可执行查询、选对来源、洗净证据、标清新鲜度、补上反证。内部固定串联5个独立技能：query-planner → source-router → evidence-cleaner → freshness-judge → count...
⭐ 0· 106·0 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 match the instructions: this is a pipeline orchestrator (query-planner → source-router → evidence-cleaner → freshness-judge → counter-evidence-hunter). It requests no binaries, no environment variables, and no install — all proportionate for an instruction-only orchestration skill. Referenced sources (e.g., 'tavily', 'bailian', 'web_search') are plausible search targets for the described use case.
Instruction Scope
SKILL.md strictly limits the skill to orchestration tasks and does not instruct reading arbitrary system files or secrets. It does instruct the agent to call five child sub-skills and to perform searches (including direct web fetchs). The workflow includes fallback chains, dynamic feedback and potential generation of additional queries. This grants the agent broad discretion over external search activity (multiple web sources and 'web_fetch_direct') but does not direct exfiltration of credentials or unrelated system data.
Install Mechanism
Instruction-only (no install spec, no code files). This is the lowest install risk — nothing will be written to disk by an installer step from this package itself.
Credentials
No required environment variables, credentials, or config paths are declared. The skill references third‑party sources but does not request API keys or secrets in its metadata, which is appropriate for a generic orchestration skill.
Persistence & Privilege
always:false and normal autonomous invocation are used. The spec describes writing pipeline_metadata, pending_actions and mentions 'local cache (memory/*_cache.md)' as a fallback — implying the orchestrator may use agent memory or writable cache. That is reasonable for a pipeline but worth reviewing in your environment to confirm where those pending_actions and pipeline_metadata are stored and who/what can read them.
Assessment
This skill appears coherent for its stated purpose, but before installing check the following: (1) child sub-skills (query-planner, source-router, evidence-cleaner, freshness-judge, counter-evidence-hunter) must exist and be trusted — inspect their SKILL.md files because the orchestration relies on them; (2) confirm how the agent performs web fetches and what connectors (tavily, bailian, web_fetch_direct) are configured — ensure those connectors don't require you to supply credentials you don't want shared and that their endpoints are vetted; (3) verify where pipeline_metadata, pending_actions and any 'local cache' are persisted and who can read them (agent memory, files, or external storage), since they may contain search queries or intermediate evidence; (4) expect this skill to issue potentially many external searches (quota/cost/privacy), so set search budgets or rate limits as needed; (5) because the skill can autonomously invoke sub-skills, review those sub-skills for any broader system access (file, network, credentials). If you cannot inspect the child skills or the connectors, consider treating the package as untrusted or run it in a limited/sandboxed agent profile.Like a lobster shell, security has layers — review code before you run it.
analysisvk97ba5865z4kgkhy26achy97mh83mbynevidencevk977ecq1sgswh83fwx0amr96m983nk2ygroundingvk977ecq1sgswh83fwx0amr96m983nk2ygrounding-pipelinevk97ba5865z4kgkhy26achy97mh83mbynlatestvk97ba5865z4kgkhy26achy97mh83mbynsearchvk97ba5865z4kgkhy26achy97mh83mbynsignal-intelligencevk97ba5865z4kgkhy26achy97mh83mbyn
License
MIT-0
Free to use, modify, and redistribute. No attribution required.