Probe First Research
v1.0.0Probe-first deep research — low-cost snippet reconnaissance before committing to full searches
⭐ 0· 37·0 current·0 all-time
by@lkm123
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
medium confidencePurpose & Capability
The name/description (probe-first research) matches the SKILL.md workflow: small number of web_search calls reading only snippets, then a user-confirmed deep phase that uses web_fetch. No environment variables, binaries, or installs are requested. One minor mismatch: the registry metadata lists no required tools, but the SKILL.md and README expect web_search/web_fetch/sessions_spawn and list allowed-tools including Bash, Read, Write, Edit — these are plausible for a research skill but broader than strictly necessary for the probe phase.
Instruction Scope
The SKILL.md is explicit about scope: Phase 1 forbids web_fetch and opening URLs, limits to 2–3 web_search calls and snippet-only reading, and defines a clear STOP POINT (user confirmation) before deeper actions. There are no instructions in the provided excerpts that direct the agent to read system files, environment variables, or exfiltrate data. Later phases (deep search, synthesis) will use web_fetch and potentially sessions_spawn as described, which is coherent with the stated purpose.
Install Mechanism
This is an instruction-only skill with no install spec and no code files to execute on install. That is the lowest-risk install model.
Credentials
The skill does not request environment variables, credentials, or access to config paths. It does request use of web_search/web_fetch and (optionally) sessions_spawn in the README/SKILL.md, which align with research functionality and are proportionate.
Persistence & Privilege
always is false, and model invocation is allowed (the platform default). The skill does not request permanent system presence or to modify other skills' configs. Multi-agent (sessions_spawn) can increase blast radius if abused, but its use is clearly described as optional for complex tasks.
Assessment
This skill appears internally consistent with its stated 'probe-first' research purpose and doesn't ask for credentials or installs, but review and consider the following before enabling: (1) Allowed-tools include Bash and file Read/Write/Edit — these allow local file access and shell execution; if you don't want those capabilities, restrict or sandbox the skill. (2) The skill expects web_search and web_fetch tools and may spawn parallel sessions (sessions_spawn) during deep work; ensure those tools are available and you accept parallel agent activity. (3) Because the skill will fetch external pages during Phase 3, consider running it in a session with limited privileges and audit the outputs/sources it cites. If you want higher assurance, inspect the full SKILL.md and README (they're included) for any additional behavior in later phases and test the skill on non-sensitive queries first.Like a lobster shell, security has layers — review code before you run it.
latestvk97fscc12fjrj04c5gqj33kx3x84tsh4
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
Runtime requirements
🔬 Clawdis