Skip to main content
Glama

omniseek_curator_act

Run source-lifecycle curation actions: submit candidates, probe evidence, decide verdicts, admit or retire sources, with safety gates enforced automatically.

Instructions

Use WHEN acting on the source-curation protocol — WRITE a source-lifecycle action (submit / probe / decide / admit / retire ...); every safety gate lives in the impl, unchanged. Pick the action with verb; each verb's REQUIRED args (see the /curator protocol):

• submit (name, urls, mode, domain, family; optional kind, regions, rationale, draft) -> add a CANDIDATE source to the admission backlog. mode ∈ STRUCTURE/UNWALL/TRANSCRIBE/RECALL/MONITOR. draft (foundry-grade) is a WORKING artifact ({"row", "fixture", "probe_summary"}) surfaced in the packet and preferred as stage_commit's ready-to-paste block. • probe (candidate_id) -> run the MECHANICAL evidence-gatherers, persist + return the packet. • wall_probe (candidate_id) -> P2 re-probe: RENDER the candidate in the network-isolated jail (egress only via the SSRF-pin proxy) so a source whose real content the plain-HTTP probe MISSED (client-rendered SPA / anti-bot / soft-login-wall) is measured on its REAL content. Eligible on a parked_p2 candidate OR an awaiting_verdict one YOU judge to be a client-rendered shell (WHEN to spend a render is your call, not an auto-gate). Surfaces content -> lands in awaiting_verdict on the rendered packet (parked_p2 revives, awaiting_verdict re-enriches); nothing -> stays put with the reason. Facts are render-derived (M7): the code never admits, only surfaces. • decide (candidate_id, decision, reasons; baseline_ref required to admit) -> record the admit/watch/reject verdict. MECHANICALLY REFUSES an admit on hard red-line / incomplete evidence / empty baseline_ref / no packet. admit -> owner_review; watch -> watching; reject -> rejected. • apply_live (candidate_id) -> ONE-TAP LIVE ADMIT (rss-safe subclass only): a REVERSIBLE overlay row + live re-register, NO git. Non-auto families are refused (use stage_commit). • rollback_live (name, family) -> full revert of a live-applied overlay row (unregister + drop). • stage_commit (candidate_id) -> ONE-TAP STAGED COMMIT for the NON-auto subclass: prepares the git commit text (does NOT apply); the operator does the git add / commit / deploy by hand. When the candidate has a foundry draft, the draft row IS the ready-to-paste block (+ a provenance line). • retire_live (name; confirm) -> ONE-TAP PRUNE (needs an existing PRUNE verdict): confirm=False previews; confirm=True writes a reversible runtime explicit_only override + stages the git commit. • rollback_retire (name) -> drop the runtime retire override so the source rejoins the fan-out. • source_verdict (name, verdict, rationale; prune_class, coverage_impact) -> record KEEP / WATCH / PRUNE for an EXISTING source. MECHANICALLY REFUSES a PRUNE the source's safety flags forbid.

Unknown verb returns an error dict listing the valid values.

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
kindNo
modeNo
nameNo
urlsNo
verbYes
draftNo
domainNo
familyNo
confirmNo
reasonsNo
regionsNo
verdictNo
decisionNo
rationaleNo
prune_classNo
baseline_refNo
candidate_idNo
coverage_impactNo

Schema Changelog

Changes observed during successful MCP inspections. Dates show when Glama detected each change.

  1. First observedv0.2.0

TDQS

A4.7/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations provided, the description carries the full burden and handles it well. It discloses side effects (adds candidate, records verdict, writes overlay row), mechanical safety gates ('MECHANICALLY REFUSES an admit on hard red-line', 'MECHANICALLY REFUSES a PRUNE'), reversibility ('REVERSIBLE overlay row', 'rollback_live'), and the fact that unknown verbs return an error dict. It also clarifies that wall_probe only surfaces render-derived facts and never admits, which prevents a dangerous misassumption.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is long, but it is a multi-verb dispatcher and nearly every line adds necessary invocation detail. The front-loaded opening states the tool's purpose immediately, and the verb-by-verb format is scannable. Some jargon and parentheticals (e.g., SSRF-pin proxy, M7, foundry-grade) add density, but they are meaningful for correct action selection rather than filler.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

This is a complex tool: 18 parameters, 10 distinct verbs, conditional requirements, state transitions, and no output schema or annotations. The description covers all verbs, their arguments, side effects, refusal conditions, and error behavior, and even describes return surfaces ('persist + return the packet', 'surfaced in the packet', 'returns an error dict'). Nothing essential for an agent to select and invoke the correct action is missing.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters5/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

The schema provides no descriptions for any of the 18 parameters, but the description compensates comprehensively by mapping each verb to its required and optional parameters: submit requires name, urls, mode, domain, family; decide requires baseline_ref to admit; retire_live needs confirm; source_verdict takes prune_class and coverage_impact. It also explains parameter semantics such as mode enum values, confirm=False preview behavior, and the draft working-artifact structure. This far exceeds what the bare schema offers.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description opens with a clear directive: use this tool when acting on the source-curation protocol and performing a WRITE source-lifecycle action. It enumerates the exact verbs (submit, probe, decide, apply_live, etc.), each with a specific resource and effect, which fully disambiguates it from read-only siblings like omniseek_curator_view. This is a specific verb+resource statement, not a tautology.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description explicitly says when to use the tool ('Use WHEN acting on the source-curation protocol — WRITE a source-lifecycle action') and gives per-verb eligibility conditions, such as wall_probe being eligible on parked_p2 or an awaiting_verdict candidate, and stage_commit being for the non-auto subclass. It does not spell out explicit alternatives like 'use omniseek_curator_view for reads,' but the WRITE qualifier and sibling context make the boundary clear.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/Battam1111/omniseek'

If you have feedback or need assistance with the MCP directory API, please join our Discord server