Skip to main content
Glama
Sign In

Hyreflow

by ai.hyreflow
OverviewDiscussionsAdmin

hyreflow_shortlist

Destructive

Bulk-screen candidates against a job description: one structured decision per candidate, ranked by score, with a tier (tier_1 | tier_2 | no_fit | unscored) and per-category scores. Takes job_spec plus either candidates (up to 1000 row objects, e.g. what a CRM/ATS search returned) or input (a ds_ dataset id, for a larger pool). Rows need work history; a pool of sourcing stubs without it is refused. The question set is generated from the job spec once and reused for every later screen of it. The call waits about 20s: a finished screen returns the top 25 ranked rows (name, row, shortlist cell); an unfinished one (including every first screen of a new job spec, which also writes its questions) returns {run_id, status, progress, poll_after_s}, and a call with only run_id returns the ranking once ready, at no charge. Every row, full profile included, is stored in the dataset at _meta.dataset_id. Resending an identical screen within 15 minutes returns the first one, charged 0. dry_run: true quotes the cost without running. Charges credits.

Input Schema

NameRequiredDescriptionDefault
inputNoA ds_ dataset id to screen instead of `candidates`.
run_idNoThe `run_id` of an earlier screen that returned unfinished. Sent alone, it returns that screen's progress or ranking. Free.
dry_runNo
job_specNoThe job description text.
min_scoreNo
questionsNoA custom question set to use instead of the generated one.
candidatesNo
must_havesNoHard requirements, each a gate: a clear no rejects the candidate, an unsure one is flagged under `unverified`. Draft them from the job spec and confirm them with the user before screening. Gate only what a work history shows (role, domain, location, stack), not language level, team size or degrees.
user_promptYesThe user's request that drove this call, verbatim as they phrased it. Recorded with the run so its trace reads as the ask followed by the calls it drove.
confirm_tokenNoThe `confirm_token` returned by a previous identical call that came back `confirmation_required`. One way to run a call above the workspace's confirmation threshold; `approved_credits` is the other.
approved_creditsNoCredits the user approved for this call after a `confirmation_required` quote. The call runs when its estimate is at or under this figure and it matches that quote.
allow_thin_profilesNo
regenerate_questionsNoReplace this job spec's stored question set.

Schema Changelog

Changes observed during successful MCP inspections.

  1. Changed2 schema fields changed
    • addedInput schema / properties / approved_credits
      Added value: +{
+  "description": "Credits the user approved for this call after a `confirmation_required` quote. The call runs when its estimate is at or under this figure and it matches that quote.",
+  "minimum": 0,
+  "type": "number"
+}
    • changedInput schema / properties / confirm_token / description
      Previous value: -"The `confirm_token` returned by a previous identical call that came back `confirmation_required`. Required to run a call above the workspace's confirmation threshold."New value: +"The `confirm_token` returned by a previous identical call that came back `confirmation_required`. One way to run a call above the workspace's confirmation threshold; `approved_credits` is the other."
  2. Changed1 schema field changed
    • changedInput schema / properties / must_haves / description
      Previous value: -"Hard requirements, each a pass/fail gate."New value: +"Hard requirements, each a gate: a clear no rejects the candidate, an unsure one is flagged under `unverified`. Draft them from the job spec and confirm them with the user before screening. Gate only what a work history shows (role, domain, location, stack), not language level, team size or degrees."
  3. Changed5 schema fields changed
    • changedInput schema / properties / confirm_token / description
      Previous value: -"Echo of the `confirm_token` a previous identical call returned with `confirmation_required`, after the user approved the quoted cost."New value: +"The `confirm_token` returned by a previous identical call that came back `confirmation_required`. Required to run a call above the workspace's confirmation threshold."
    • changedInput schema / properties / must_haves / description
      Previous value: -"Hard requirements, each a pass/fail gate. Confirm them with the user first."New value: +"Hard requirements, each a pass/fail gate."
    • changedInput schema / properties / questions / description
      Previous value: -"Your own question set instead of the generated one — shapes in shortlist.md."New value: +"A custom question set to use instead of the generated one."
    • changedInput schema / properties / run_id / description
      Previous value: -"Poll a screen an earlier call handed back unfinished — send only this (and user_prompt). Free."New value: +"The `run_id` of an earlier screen that returned unfinished. Sent alone, it returns that screen's progress or ranking. Free."
    • changedInput schema / properties / user_prompt / description
      Previous value: -"The user's request that drove this call, verbatim as they phrased it. Never a paraphrase, and never ask the user for it."New value: +"The user's request that drove this call, verbatim as they phrased it. Recorded with the run so its trace reads as the ask followed by the calls it drove."
  4. Changed2 schema fields changed
    • addedInput schema / properties / run_id
      Added value: +{
+  "description": "Poll a screen an earlier call handed back unfinished — send only this (and user_prompt). Free.",
+  "type": "string"
+}
    • changedInput schema / required
      Previous value: -[
-  "job_spec",
-  "user_prompt"
-]New value: +[
+  "user_prompt"
+]
  5. Changed2 schema fields changed
    • removedInput schema / properties / async
      Removed value: -{
-  "type": "boolean"
-}
    • addedInput schema / properties / regenerate_questions
      Added value: +{
+  "description": "Replace this job spec's stored question set.",
+  "type": "boolean"
+}
  6. Added

TDQS

A4.5/5.0
Behavior5/5

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

With annotations only declaring destructive/openWorld, the description carries real operational weight: it discloses credit charging, dry-run quoting, the 15-minute idempotency window returning the first run at zero cost, the ~20s wait and the run_id polling contract that is free, plus the fact that questions and full row profiles are persisted to a dataset. It also flags the refusal behavior for thin sourcing stubs.

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 core capability and input choice are front-loaded, and nearly every sentence carries operative detail. The mid-section mixing polling, dataset storage, dedupe and charging is dense and slightly jumbled, but nothing is superfluous.

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

Completeness4/5

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

For a 13-parameter, destructive, credit-charging tool with no output schema, the description does the heavy lifting: it explains both return shapes (ranked top-25 rows vs {run_id, status, progress, poll_after_s}) and the persisted dataset. Gaps remain around `min_score`, `allow_thin_profiles`, and `regenerate_questions` semantics.

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

Parameters4/5

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

Schema coverage is 69%, and the description adds genuine meaning for the main parameters: `job_spec`, `candidates` (row objects from a CRM/ATS search, needing work history), `input` (ds_ dataset id), `run_id`, `dry_run`, and the generated-and-reused `questions`. It leaves `min_score`, `allow_thin_profiles`, and `regenerate_questions` unaddressed, which the schema does only partially.

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 specific verb and resource ('Bulk-screen candidates against a job description') and immediately states the output shape: one structured decision per candidate, ranked, with tier and per-category scores. No sibling in the list does candidate screening, so an agent can identify this tool without reading the schema.

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?

It clearly distinguishes the two input paths (inline `candidates` up to 1000 rows vs a `ds_` dataset id via `input` for a larger pool), and explains the polling path via `run_id`. It does not address when to prefer a sibling like `hyreflow_enrich_job` or `hyreflow_dataset_read` instead, so routing between tools remains inferred.

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

All Hyreflow toolsInspect
Try in Browser
One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Glama MCP Gateway

Add one secure layer between your agents and this server.

Resources