/comp-benchmark

Purpose

/comp-benchmark

What it IS: a structured first pass that produces a market reference grid (10th / 25th / 50th / 75th / 90th percentile base pay bands) a human compensation analyst or HR leader can review, interrogate, and own, with every source, SOC code, adjustment factor, and data quality flag surfaced explicitly.

What it is NOT: a pay equity audit (see hard-scoped-out section below and future

/pay-equity-audit

This skill is governed by

hr-ai-governance

When to Use

Invoke

/comp-benchmark

• Open a new role and build a market reference grid before setting an internal comp band
• Refresh an existing comp band during annual review using current public data
• Check jurisdictional exposure when expanding a role into a new state / city / country with pay-transparency rules
• Provide competitive-offer context (public-data reference) for an open requisition
• Feed a reference grid into

  /job-description-generator

  for a jurisdiction that requires pay-range-on-posting

When NOT to Use

Do NOT use

/comp-benchmark

• You need a pay equity audit — out of scope, HARD block; use

  /pay-equity-audit

  (future phase, requires attorney-client privilege and documented

  @employment-counsel

  engagement)
• You need an individual offer (base + bonus + equity + sign-on) — that's a hiring manager + HR leadership decision involving internal equity, capability differentiation, and budget, not a market data pull
• You need internal equity analysis (existing team bands vs. new hire) — out of scope, separate workflow with confidential HR leadership
• You need retention risk assessment for a specific individual — out of scope, confidential HR
• You need variable compensation modeling (bonus targets, equity refresh math, commission plans) — out of scope, separate future skill
• You need market data from a vendor survey (Radford, Mercer, Willis Towers Watson, Payscale, Salary.com, Glassdoor, Levels.fyi, H1B wage DB) — deliberately excluded; the skill's entire point is public-source-only discipline
• You need a legal opinion on whether a specific pay-transparency obligation applies — use

  @employment-counsel

The skill deliberately sits UPSTREAM of

/job-description-generator

Source Whitelist (MANDATORY)

This is the complete list of sources

/comp-benchmark

Explicitly EXCLUDED (any appearance in a run is a quality gate failure):

Radford, Mercer, Willis Towers Watson (WTW), Payscale, Salary.com, Glassdoor, Levels.fyi, H1B Wage Database, Comparably, Indeed Salary, Built In, LinkedIn Salary Insights, any other vendor compensation survey, any scraped job board data, any "what I've seen at similar companies" recollection, any "industry knowledge" not tied to one of the eight whitelisted sources.

The H1B Wage Database is specifically called out because it is free and public but has known systematic bias (selection effect: only visa-sponsored roles, skewed toward tech, skewed toward larger employers) that makes it unsafe as a general reference. It is deliberately excluded.

Every source used in a run MUST be cited by name, URL, date accessed, SOC code queried, and geography queried in the methodology worksheet.

Required Inputs

The skill MUST collect the following before producing output. If any required input is missing, it asks the user rather than guessing.

US-DE

US-NYC

US-CA

US-CO

US-WA

US-IL

EU-DE

UK

IL

$130,000-$170,000

Jurisdiction is never defaulted. If the user does not supply it, the skill asks. The jurisdiction determines pay-transparency disclosure obligations, applicable runtime warnings, and audit log retention per

hr-ai-governance

SOC code is never assumed from title. Role → SOC mapping is Method Step 1; the skill walks the O*NET crosswalk deterministically where possible and flags ambiguity as a data quality finding when the role does not have a clean SOC.

Output Structure

Every

/comp-benchmark

sensitive-skill-guardrails.md

1. Disclaimer + UPL Guardrail Block (top)

> ⚠️ **Not HR or employment-law advice.** This output is a drafting and triage aid generated by a product-organization skill, not HR counsel, compensation counsel, or employment counsel. No attorney-client relationship is created by its production or use. Jurisdiction-specific questions on pay-transparency obligations, pay equity compliance, protected-class pay analysis, or enforceability of compensation commitments require review by a licensed employment attorney in the relevant jurisdiction and by a qualified compensation professional. Compensation decisions involving individual offers, internal equity, pay equity, or retention require HR leadership review. Do not rely on this output as the sole basis for setting any pay band, making any individual offer, or responding to any pay-transparency obligation.
>
> **Jurisdiction Assumed:** {jurisdiction from the required input}. If your jurisdiction differs, treat every finding below as a hypothesis to verify with local counsel.

2. Benchmark Metadata block

A labeled block under the disclaimer with:

• Role: title as input
• SOC Code Assigned: primary SOC code chosen in Method Step 1 (e.g.,

  11-2021 Marketing Managers

  ) + any candidate SOCs considered
• SOC Assignment Method:

  deterministic

  /

  O*NET crosswalk

  /

  ambiguous-analyst-choice

• Jurisdiction: as input (no default)
• Target Geography: MSA / state / national / remote-national
• Seniority Level: as input
• Pay Transparency Required:

  yes

  /

  no

  (derived from Section 8.1 jurisdiction matrix + Section 5 of this skill)
• Pack Inheritance:

  hr-ai-governance v{version}

  (read at runtime)
• Generation Timestamp: ISO 8601 UTC
• Governance Pack Version:

  hr-ai-governance v1.0.1+

• Data Freshness Window: youngest source date and oldest source date across all queried sources

3. The Reference Grid

A table with the stated SOC code and target geography, reporting base pay only (no variable comp):

Every value in the grid is the skill's structured extraction from whitelisted public sources. The skill does NOT invent values, does NOT interpolate across SOCs without documentation, and does NOT apply "professional judgment" adjustments. If a percentile cannot be sourced, the row is populated with

[cannot-source]

Variable compensation (bonus, equity, commission) is EXPLICITLY excluded from the grid. The public data sources do not cleanly report variable comp; attempting to benchmark variable comp from public sources would produce misleading numbers. Variable comp is a separate future skill.

4.

## Methodology Worksheet

A structured worksheet documenting every step the skill took to produce the grid. This is the audit-trail output — it is the most important section for a human reviewer.

Fields:

• SOC Assignment:
  • Primary SOC chosen: code + title
  • Alternate SOCs considered: list
  • Rationale for choice: one-paragraph explanation tied to role description
  • Ambiguity flag:

    clean

    /

    ambiguous-primary-chosen

    /

    no-clean-match

• Sources Queried (one row per source):
  • Source name (from whitelist)
  • URL accessed
  • Date accessed
  • SOC code queried
  • Geography queried
  • Data year (source vintage)
  • Result summary
• Geography Adjustment:
  • Was adjustment applied: yes / no
  • If yes: adjustment factor + source (BLS ECI metro ratio or ACS-derived metro wage ratio)
  • Documented reasoning
• Seniority Adjustment:
  • SOC-wide wage distribution captures an occupation, not a seniority level. Seniority within SOC requires a differentiation factor. The skill documents how the seniority adjustment was derived (e.g., "percentile position within SOC: Lead corresponds to 75th+ percentile within management SOC codes per BLS OEWS distribution notes") and flags the reliability of the adjustment
  • Method: percentile-banding / direct-match-from-pay-transparency-disclosures / cannot-derive
• Assumptions: numbered list of every assumption the skill made to produce the grid
• Excluded Sources: confirmed list of which whitelisted sources were NOT used for this run and why (e.g., "OPM: role is private-sector, not federal-equivalent")

5.

## Jurisdiction Compliance Report

• Jurisdiction: as input
• Pay Transparency Framework: the specific rule(s) in force (e.g., "US-NYC: NYC DCWP pay-transparency law — range on posting required"; "US-CO: Colorado Equal Pay for Equal Work Act + Colorado AI Act SB24-205 deployer obligations; range on posting required; SB24-205 applies if AI substantial factor in consequential decision"; "US-DE: no pay-transparency rule in force as of 2026-04-11")
• Disclosure Obligation:

  range-on-posting

  /

  range-on-request

  /

  range-at-interview

  /

  not-applicable

• Illinois HB 3773 trigger check: applied if jurisdiction includes IL, flags adverse-impact check obligation per

  hr-ai-governance

  Section 2
• Colorado AI Act trigger check: applied if jurisdiction includes CO, flags deployer obligations per

  hr-ai-governance

  F11 / Section 8.1
• EU pay-transparency directive (2023/970) check: applied if jurisdiction is EU, flags member-state transposition status and applicable date
• Compliance Posture:

  ready-to-disclose

  /

  requires-band-construction-before-posting

  /

  not-applicable

  /

  consult-employment-counsel

6.

## Data Quality Flags

Numbered findings covering data quality and reliability. Each flag gets a row with:

• Finding #
• Category —

  thin-SOC-coverage

  /

  outdated-source

  /

  geography-extrapolation

  /

  seniority-adjustment-low-reliability

  /

  SOC-ambiguity

  /

  whitelist-gap

  /

  freshness-warning

• What: specific issue
• Why it matters: impact on the grid's reliability
• Severity: P0 / P1 / P2
• Verdict:

  address

  /

  accept-with-risk

  /

  cannot-address-within-whitelist

• Suggested next step: what a human reviewer should do

7.

## Findings

(substantive)

Numbered findings beyond data quality. Examples:

• "SOC coverage thin for this geography (ACS 5-year not available at this MSA level); national BLS OEWS used with ECI metro ratio; reliability: moderate"
• "Latest source vintage is 18 months old for seniority-matched percentile; data within freshness window but nearing edge"
• "Jurisdiction requires pay-range disclosure at posting; grid supports compliant range construction at P25-P75"
• "Role title 'Customer Success Lead' does not have a clean SOC match; three candidate SOCs considered (see methodology worksheet); primary choice documented with rationale"

8.

## Reviewer Checklist

Eight mandatory sign-off items before the grid is acted upon:

• Jurisdiction confirmed against role posting location and the employer's presence
• SOC code choice reviewed and accepted by compensation analyst
• Geography adjustment method reviewed and accepted (or corrected)
• Seniority adjustment method reviewed and accepted (or corrected)
• Every P0 data quality flag addressed or explicitly accepted-with-risk
• Jurisdiction compliance posture reviewed; employment counsel engaged if required
• No vendor-survey leakage in sources (grep check passed)
• Pay equity analysis explicitly confirmed out of scope (this skill does not produce a pay equity audit)

9.

## Cannot Assess Without

Minimum 7 items. This section makes scope explicit:

• Internal equity — existing team comp bands. Separate workflow; confidential HR leadership.
• Individual capability differentiation — hiring manager judgment on a specific candidate's value to the role. Not a market data question.
• Budget constraints — finance input; depends on headcount plan, runway, organizational priorities. Out of scope.
• Variable compensation modeling — bonus targets, equity refresh math, commission plans. Separate future skill; public data does not support clean variable comp benchmarking.
• Retention risk for a specific individual — confidential HR leadership; depends on attrition signals, manager feedback, career conversations.
• Pay equity compliance — requires

  /pay-equity-audit

  (future phase) under attorney-client privilege with documented

  @employment-counsel

  engagement. Hard-scoped out of this skill. See the Pay Equity Rejection section below.
• Non-whitelisted vendor survey data — Radford, Mercer, WTW, Payscale, Salary.com, Glassdoor, Levels.fyi, H1B Wage DB. Deliberately excluded by source whitelist discipline. If the user wants vendor data, they must obtain it separately and this skill is not the tool.
• Enforceability of compensation commitments — offer letters, employment agreements, equity vesting clauses. Requires

  @contracts-counsel

  +

  @employment-counsel

  .

10. Audit Log Entry

Emitted per

hr-ai-governance

• run_id

  ,

  timestamp

  ,

  skill_name

  ,

  skill_version

  ,

  governance_pack_version

• jurisdiction

• inputs_hash

  — SHA-256 of the canonicalized role-input object (role, jurisdiction, seniority, geography, SOC choice). NEVER raw candidate data — this skill doesn't see candidate data.

• input_type: comp_record

• proxies_detected

  — from Method Step 3 proxy scan; if the combination of geography + title + industry + company size triggers a proxy flag per

  hr-ai-governance

  Section 3.2, each flagged proxy becomes an entry

• redactions_applied

  — N/A for this skill (compensation data inputs are not candidate data)

• hitl_reviewer

  — authenticated identity of the compensation analyst or HR leader who will review

• hitl_decision

  ,

  hitl_timestamp

  ,

  hitl_rationale

  — populated at HITL gate

• downstream_action

  — e.g.,

  band-approved

  ,

  band-rejected

  ,

  band-sent-to-jd-generator

• signoff_path: standard

• retention_expiry

  — computed from jurisdiction per Section 4.3

The skill refuses to run if the deployer's auth system cannot supply an authenticated

hitl_reviewer

Method

The skill's benchmark flow is a 7-step pipeline. Order matters because later steps depend on earlier ones.

Step 1 — Map Role to SOC Code

The SOC (Standard Occupational Classification, BLS-maintained) is the neutral, machine-consumable, jurisdiction-neutral taxonomy the skill uses to anchor every benchmark run. Role → SOC mapping is the most consequential step because every downstream query uses the SOC code.

Procedure:

1. Read the role title and any description text the user supplied
2. Query O*NET OnLine crosswalk for the title → SOC candidate list
3. If exactly one clean SOC matches, set primary SOC and set

   SOC Assignment Method = deterministic

4. If multiple SOCs are plausible, enumerate candidates (up to 5), assess each against the role description, and choose the primary SOC with a one-paragraph rationale; set

   SOC Assignment Method = O*NET crosswalk

5. If no clean SOC exists (novel role, hybrid function, emerging job category), document the three closest candidate SOCs, pick the primary with rationale, set

   SOC Assignment Method = ambiguous-analyst-choice

   , and flag a

   SOC-ambiguity

   finding at P1 severity
6. For ambiguous cases, consultation with

   @recruiter

   is available (Pattern 1 Consultation) — the recruiter sees role titles every day and has practical SOC-mapping intuition

The SOC choice and the alternates considered are ALWAYS documented in the methodology worksheet. There is no "silent SOC choice."

Step 2 — Query BLS OEWS for Primary Grid

Using the primary SOC code from Step 1 and the target geography from the inputs:

1. Query BLS OEWS at the finest geography granularity that has data for this SOC (MSA → state → national fallback)
2. Extract the 10th / 25th / 50th / 75th / 90th percentile annual wages
3. Record the OEWS data year (vintage)
4. Record geography granularity actually used (e.g., "MSA data available" vs. "fell back to state data" vs. "fell back to national data")

If OEWS has no data for this SOC at any geography, the skill blocks and routes to Step 3 for secondary sources before continuing.

Step 3 — Fill Gaps from Secondary Whitelist Sources + Proxy Scan

If OEWS coverage is thin or absent:

1. Query ACS for the target metro / county with the primary SOC
2. Query QCEW for industry + geography context
3. If role is federal-equivalent, query OPM GS tables for the target locality pay area
4. If jurisdiction has pay-transparency rules, query state/city pay-transparency job board disclosures and municipal pay scales for directly observed pay ranges in the target metro
5. Run the

   hr-ai-governance

   Section 3.2 proxy scanner on the input combination: does (geography + title + industry + company size) create a proxy for age / race / gender / national origin? E.g., a highly gender-skewed SOC in a racially-skewed metro with a small company filter can surface proxies that need to be controlled. Each detected proxy becomes a

   proxies_detected

   entry in the audit log and a finding in the Data Quality Flags section.

All secondary source queries are documented in the methodology worksheet row-by-row.

Step 4 — Apply Geography Adjustment (if needed)

If OEWS or ACS has the SOC at the target geography at clean granularity, no geography adjustment is applied. Otherwise:

1. Query BLS ECI metro series for the target metro — ECI gives a cost-of-labor index
2. If ECI is not available for the target metro, derive an ACS metro wage ratio (target metro average wage / reference geography average wage)
3. Apply the adjustment factor to the reference grid percentiles
4. Document the adjustment factor, its source, and the reasoning in the methodology worksheet
5. The user can override the adjustment with explicit reasoning; the override is recorded in the audit log

Geography adjustment is one of the two most reviewer-scrutinized parts of the grid. Every adjustment is documented, not silent.

Step 5 — Apply Seniority Adjustment

SOC codes capture an occupation, not a seniority level within the occupation. A "Marketing Manager" SOC includes everyone from entry-level manager to VP-level manager; the BLS OEWS percentile distribution captures that full range. Seniority within a SOC requires a differentiation factor.

Procedure:

1. Read the stated seniority level
2. Map seniority to a percentile position within the SOC wage distribution:
   • Junior / IC2 → 10th-25th percentile range
   • IC3 / Mid → 25th-50th percentile range
   • IC4 / Senior → 50th-75th percentile range
   • IC5 / Lead / Staff → 75th-90th percentile range
   • Principal / Director+ → 90th+ percentile range (and note that BLS reports "90th percentile" as a single value, not a distribution above it)
3. If pay-transparency job board disclosures provide direct seniority-labeled pay observations at the target geography, those override the percentile-banding method (direct evidence wins)
4. Document the seniority adjustment method and reliability level in the methodology worksheet
5. If the seniority adjustment cannot be defensibly derived (e.g., the SOC is too broad, the role's seniority doesn't map cleanly to a percentile band, the direct evidence doesn't exist), flag

   seniority-adjustment-low-reliability

   at P1

Seniority adjustment reliability is the second most reviewer-scrutinized part of the grid. The skill is conservative here: low reliability is flagged, not smoothed over.

Step 6 — Run Jurisdiction Compliance Check

Per the jurisdiction loaded in the input:

1. Look up the row in

   hr-ai-governance

   Section 8.1
2. Determine the pay-transparency obligation (range-on-posting / range-on-request / range-at-interview / not-applicable)
3. If the jurisdiction is US-IL, trigger the HB 3773 adverse-impact check flag (per pack Section 2) as the grid may feed a JD or hiring pipeline
4. If the jurisdiction is US-CO, trigger the Colorado AI Act deployer obligations flag (pre-use notice, right to correct, annual impact assessment, 90-day algorithmic-discrimination notification to Colorado AG) per pack F11
5. If the jurisdiction is EU, check if the target member state has transposed Directive 2023/970 (transposition deadline was June 7, 2026, varying by member state) and flag applicable obligations
6. If the jurisdiction is not in the matrix, block and route to

   @employment-counsel

The compliance report surfaces obligations; the skill does not execute them. The deployer / HR leader is responsible for compliance actions.

Step 7 — Generate Reference Grid + Methodology Worksheet

Assemble the grid, the methodology worksheet, the jurisdiction compliance report, and the data quality flags into the output structure defined above.

Run the 9-check quality gate (below) before emitting.

Pay Equity Rejection (Hard Scope-Out)

Pay equity analysis is out of scope for

/comp-benchmark

If a user asks the skill to perform pay equity analysis — including but not limited to: "is our pay equitable?", "are women at our company paid fairly?", "do we have a gender pay gap?", "check for racial pay disparities in engineering," "audit our comp bands for protected-class pay gaps," "is our pay band fair under the Equal Pay Act?" — the skill:

1. Refuses to proceed

2. Returns a routing note:

   "Pay equity analysis is out of scope for

   /comp-benchmark

   . Pay equity is legally and methodologically distinct from market benchmarking. It requires:

   • Attorney-client privilege (so the analysis and its findings are protected from disclosure in litigation)
   • Documented

     @employment-counsel

     engagement (so the privilege attaches)
   • A different methodology: regression analysis on internal compensation data controlling for legitimate factors (tenure, performance, role, geography), with sensitivity analysis for protected-class variables
   • A different data source: internal compensation records, not public BLS data
   • A different audit log posture: privileged, attorney-work-product

   Use

   /pay-equity-audit

   (future phase — not yet available) and engage

   @employment-counsel

   BEFORE running any pay equity analysis. Running an unprivileged pay equity analysis can create discoverable evidence in litigation, which is the precise outcome the privileged path is designed to prevent."

3. Logs the rejection in the audit log as

   downstream_action: pay-equity-refused-scope

4. Does NOT produce a grid or any output that could be repurposed as pay equity analysis

The rejection is mandatory. There is no "express mode" that skips it. The reason is liability: a market benchmark grid that gets used in litigation as evidence of pay equity analysis is precisely the discoverable-exposure pattern the privileged path prevents. The skill defends the boundary.

Quality Gates

The skill performs a 9-item self-check BEFORE emitting output. If any check fails, the output does not publish — it produces a structural finding instead, asking the user to fix the gap.

1. SOC code assigned and justified — every run has a primary SOC code with a documented rationale in the methodology worksheet
2. All sources in whitelist (grep-able) — every source cited in the methodology worksheet matches one of the eight whitelisted sources exactly; any unknown source is a gate failure
3. At least one BLS source present — BLS OEWS or QCEW must anchor the grid; a run that relies only on secondary sources (ACS, OPM, pay-transparency disclosures, municipal pay) without any BLS anchor is a gate failure
4. Data freshness: every source dated within 24 months — sources older than 24 months are flagged as a

   freshness-warning

   ; sources older than 36 months are a gate failure unless explicitly accepted-with-risk by the reviewer
5. Geography adjustment documented if applied — if any adjustment factor was applied, its source, value, and reasoning must appear in the methodology worksheet; silent adjustments are a gate failure
6. Seniority adjustment documented with method — the method used to derive the seniority position within the SOC is documented; "analyst judgment" without a method reference is a gate failure
7. Jurisdiction compliance surfaced — the jurisdiction compliance report is present and the pay-transparency obligation is named (even if "not-applicable")
8. No vendor-source leakage — grep check for Radford, Mercer, Willis Towers Watson, WTW, Payscale, Salary.com, Glassdoor, Levels.fyi, H1B wage database, Comparably, Indeed Salary, Built In, LinkedIn Salary. Any hit is a gate failure.
9. Pay equity analysis explicitly rejected if requested — if the user's input asked for pay equity, the Pay Equity Rejection above MUST have fired and the skill MUST NOT have produced a grid

These 9 checks are the quality gates. They run every invocation. There is no "express mode" that skips them.

Pack Inheritance

This skill inherits the following from

hr-ai-governance

inputs_hash

jurisdiction

hitl_reviewer

signoff_path

retention_expiry

input_type: comp_record

/resume-summarizer

/job-description-generator

This skill is NOT an AEDT. It produces a market reference grid, not a candidate decision. The AEDT wall from

hr-ai-governance

Delegation Patterns Available

Default: Pattern 1 Consultation

When a specific element of the grid needs specialist input, the skill spawns a consultation per

delegation-protocol.md

Consultations are attributed in the Data Quality Flags or Findings sections: "I consulted

@recruiter

Adversarial Review

NOT applicable at this version. Adversarial Review (Pattern 5) is reserved for near-final deliverables with high-stakes, uncapped exposure (enterprise contracts, M&A documents, pricing commitments). A compensation reference grid is neither near-final nor uncapped-exposure at the generation stage; it is a draft going to a human compensation analyst who will review, interrogate, and potentially reject.

If a specific benchmark becomes high-stakes (e.g., a public-company executive comp disclosure, a class-action pay-transparency matter, a SEC-filed NEO compensation benchmark), the requester should use a different tool and engage

@employment-counsel

ROI Framing

ROI for

/comp-benchmark

HR-analytics blended rate: $200/hr (senior compensation analyst or HR-analytics specialist rate; higher than the standard HR $150/hr because the role combines BLS data literacy, jurisdictional compliance, and SOC mapping expertise).

Time-saved baseline: a careful, public-source-disciplined compensation reference grid with SOC mapping, geography adjustment documentation, seniority adjustment methodology, jurisdiction compliance check, and data quality flags is ~3-4 hours of manual drafting and triage time for a compensation analyst working with a hiring manager. Simpler refreshes (same SOC, same geography, updated BLS OEWS vintage) are ~1-1.5 hours baseline. Complex roles (ambiguous SOC, multi-jurisdiction posting, thin public data requiring multiple secondary source gap-fills) are 5+ hours baseline.

The ROI tracks ONLY the time the skill saves on drafting and triage of the structured artifact, not the substantive compensation review time. The substantive review by the compensation analyst, HR leader, and (where needed) employment counsel still happens in full.

Example ROI line for a standard benchmark run:

⏱️ ~3.5 hrs saved on drafting and triage in 55s, 18k tkns ~$1.1 cost, Value ~$700

Related Skills + Hand-off

Input

• /job-description-generator

  (bi-directional) — when a hiring manager runs comp-benchmark first to get a market grid, the grid output feeds JD-generator as the comp band input for pay-transparency jurisdictions. Conversely, when JD-generator runs first and needs a comp band, it can invoke comp-benchmark with the role + jurisdiction from its own inputs. The two skills are co-owned in spirit (JD-generator has compensation-analyst as co-owner for the Comp Band Alignment Report; comp-benchmark has job-description-generator as a consumer for the grid → JD pay range hand-off).

Output feeds

• /job-description-generator

  — comp band for pay-transparency-required JDs; the grid's 25th-75th percentile range typically seeds the JD's compensation section

• /interview-guide

  (downstream) — seniority alignment check (does the stated seniority match what the SOC + comp band implies?)

Future / deferred

• /pay-equity-audit

  — HARD scoped out. Different privilege posture (attorney-client privileged), different methodology (regression on internal data), different data source (internal comp records not public BLS), different audit log posture. Requires

  @employment-counsel

  engagement before running. Not in Phase 4A; not in this skill.
• Variable comp benchmark skill — future, separate skill. Public data does not support clean variable comp benchmarking; attempting to benchmark bonus / equity / commission from BLS data would produce misleading numbers.

Pack

• hr-ai-governance

  v1.0.1 — every run inherits proxy register (3.2), audit log schema (4.1), jurisdiction matrix (8.1), HITL enforcement (10), review cadence (11.1)

Attribution and Maintenance

Owner: 💵 Compensation Analyst. The skill's SOC mapping, source whitelist discipline, BLS query logic, geography and seniority adjustment methodology, and pay equity scope-out are Compensation Analyst's accountability.

Consumers (skills / gateways that invoke this skill):

• ext-hr

  — HR team gateway, primary user

• job-description-generator

  — bi-directional; comp band input/output hand-off

• interview-guide

  — downstream; seniority alignment check

New consumers require a frontmatter update and a one-line note in the consuming gateway's or skill's dependency list.

Authoring: First-principles. This skill was authored from scratch during Phase 4A as the second HR skill under the

hr-ai-governance

/job-description-generator

Dependency on the pack: The skill reads from

hr-ai-governance

Updates: Via the two-pass publication gate defined in

sensitive-skill-guardrails.md

/resume-summarizer

/job-description-generator

Minor edits (typos, formatting, jurisdiction matrix syncs with pack updates) can bypass Pass 2. Any edit touching: the source whitelist (adding or removing a source), the SOC mapping method, the geography adjustment methodology, the seniority adjustment methodology, the pay equity rejection boundary, or the HITL gate placement — requires a full Pass 2 substantive review by Employment Counsel.

Changelog: Maintained at the bottom of this file.

Example Invocation

User: /comp-benchmark --role "Customer Success Lead" --jurisdiction US-DE --seniority Lead --geography "remote-national, timezone overlap IL + US-Eastern"

/comp-benchmark v1.0.0 — loading:
  - Jurisdiction: US-DE (no pay-transparency rule; grid produced for reference only)
  - Governance pack: hr-ai-governance v1.0.1
  - Source whitelist: BLS OEWS, BLS QCEW, BLS ECI, OPM, O*NET, US Census ACS, state/city pay-transparency disclosures, municipal pay scales
  - Seniority: Lead (75th-90th percentile within SOC distribution)
  - Target geography: remote-national, coastal-metro-adjustment assumed

Running pipeline:
  Step 1: SOC mapping — role is ambiguous; 3 candidate SOCs considered (11-2021, 13-1111, 43-1011); primary = 11-2021 Marketing Managers with documented rationale; SOC-ambiguity flag raised (P1)
  Step 2: BLS OEWS query — national data retrieved for 11-2021 at 10th/25th/50th/75th/90th percentile
  Step 3: Secondary sources — ACS metro wage ratios retrieved for reference metros; pay-transparency disclosures sampled for CSM Lead postings in NY/CO/CA; proxy scan run (no flags)
  Step 4: Geography adjustment — national → coastal-metro adjustment applied using BLS ECI; +8% factor documented
  Step 5: Seniority adjustment — Lead mapped to 75th-90th percentile band with percentile-banding method; seniority-adjustment-reliability: moderate (no direct seniority-labeled observations for this exact role + geography)
  Step 6: Jurisdiction compliance — US-DE: no pay-transparency rule; compliance posture: not-applicable
  Step 7: Grid + methodology worksheet + jurisdiction report + data quality flags assembled

Producing output at:
  Legionis/Product/comp-benchmark-birth-test-2026-04-11.md

9/9 quality gates passed (with SOC-ambiguity flag surfaced for reviewer decision).

⏱️ ~3.5 hrs saved on drafting and triage in 55s, 18k tkns ~$1.1 cost, Value ~$700

Changelog

• 1.0.0 (2026-04-11) — Initial authoring. First-principles during Phase 4A as the second HR skill under the

  hr-ai-governance

  pack (after

  /job-description-generator

  ). Owned by 💵 Compensation Analyst. Source whitelist constructed directly from the execution plan's Round 2 constraints following the Employment Counsel near-block on vendor-data sourcing. SOC mapping methodology sourced from BLS SOC system + O*NET OnLine public crosswalk; no vendor taxonomy used. Pay equity analysis hard-scoped out with mandatory rejection routine. Subsequent-similar 72-hour Pass 2 review under Employment Counsel SLA (the pack carries first-of-type status;

  /resume-summarizer

  carries first-of-type for skills;

  /job-description-generator

  established subsequent-similar precedent). Scaffolding review by 📋 Director of HR. Birth-tested against the same Legionis Customer Success Lead role used for

  /job-description-generator

  birth test, with explicit SOC ambiguity documented (see

  Legionis/Product/comp-benchmark-birth-test-2026-04-11.md

  ).