Audit Scoring Stinger
You are equipping audit-scoring-wasp-drone, part of the Website Auditor by Legion Code
Inc. plugin. Full scope and acceptance criteria: prd-020-audit-scoring.
This is one of the most load-bearing pairs in the plugin: wave W7, sync, needs every
applicable Wave-5/W6 Drone's findings before it can run at all, and its output (the populated
scoring/audit-scorecard.xlsx) is what audit-reporting-wasp-drone renders into both
customer- and auditor-facing deliverables in wave W8. Every factual claim below traces to a
downloaded primary source in references/research/raw/, or is explicitly flagged as this
Stinger's own engineering design where the research archive itself says the design is not
directly sourced - see guides/03-nesting-structure-design.md for the full accounting.
Purpose
Roll every leaf finding (0-6 score, evidence pointer, one-line justification) up through
sub-audit, category, and final scores using N/A-aware masked-SUMPRODUCT formulas at every
level; apply the critical-security-override (any Security leaf scored 1 caps the final grade
at C); and populate the branded, named-range-driven XLSX scorecard, per build plan sections
4.1-4.4 and prd-020's acceptance criteria.
When to use this skill
- Wave W7 of every audit run, after every applicable Wave-5/W6 Drone has written its findings
(
03-seo/ through 12-ecommerce/, per the shared workspace contract).
- Verifying a final letter grade or the security-override cap by hand against the workbook's
own formulas (e.g. during a QA pass on this pair, or when a client questions a grade).
- Retuning category or sub-audit weights via the
Rubric sheet's named ranges for one
engagement, without editing any formula (guides/04-retuning-weights.md).
- Validating an incoming leaf finding against
references/templates/leaf-finding.schema.json
and deciding whether to score it or reject it back to its originating Drone
(guides/05-rejecting-a-leaf-finding.md).
- Regenerating the branded XLSX template itself after a design change to the category/
sub-audit weighting structure (
references/scripts/generate-scorecard-xlsx.py).
When not to use this skill
- To re-score, second-guess, or reinterpret an upstream Drone's own leaf finding. That is the
originating Drone's domain judgement, not this Stinger's arithmetic. A malformed or
unevidenced finding is rejected back to its origin, never silently corrected here
(
guides/05-rejecting-a-leaf-finding.md).
- To generate the customer- or auditor-facing narrative reports. That is
audit-reporting-wasp-drone / audit-reporting-stinger's job in wave W8, reading this
pair's output as its own input.
- To crawl, capture evidence, or produce a finding from scratch. This pair only consumes
already-produced leaf findings; it has no domain expertise in security, SEO, accessibility,
or any of the other audited domains.
- To determine category weights or the override rule from first principles. Both are binding
product requirements straight from
plan/website-auditor-build-plan.md section 4.2/4.3 and
prd-020's acceptance criteria (AC-2, AC-3), not this Stinger's discretion to redesign; only
the numeric VALUES in the named ranges are meant to be retuned per engagement.
Procedure
- Copy the template. Copy
references/templates/website-audit-scorecard-template.xlsx
to the run's scoring/audit-scorecard.xlsx (build plan section 3's shared workspace path).
Never edit the template in place; never hand-edit a formula cell in the copy either
(guides/04-retuning-weights.md section 6).
- Read every applicable category folder's output (
03-seo/ through 12-ecommerce/) and
_shared/evidence-index.md, per the shared workspace contract in prd-020.
- Validate each candidate leaf finding against
references/templates/leaf-finding.schema.json. A finding that fails validation, is a
malformed boolean checkpoint, or targets an unknown category/sub-audit coordinate is
rejected back to its originating_bee per guides/05-rejecting-a-leaf-finding.md - never
scored anyway.
- Transcribe every valid leaf finding into its row on the
Scorecard sheet (score,
evidence pointer, justification - a direct write, never a formula edit). Evidence and
justification text comes from third-party pages: write it as text, and prefix any value that
starts with =, +, -, @, a tab or a carriage return with ' so no cell becomes a formula.
- Recalculate (open in Excel/LibreOffice, or force a headless recalculation pass) and
read back the rollups: sub-audit (
guides/01-rollup-procedure.md section 1-2), category
(section 3), and final (section 4). The critical-security-override
(guides/02-critical-security-override.md) resolves automatically on the
Executive Scorecard sheet from the same recalculation - no separate step is needed.
- Verify, don't just trust. Spot-check at least one sub-audit rollup and the final letter
grade by hand against the raw leaf scores before treating the run as complete, per
guides/01-rollup-procedure.md section 5's verification method. Cross-check the whole
rollup with shared/scripts/score-rollup.py (see references/scripts/README.md). The
script and the workbook apply the same N/A rule at every level, so they must agree
exactly (an N/A sub-audit or category it lists shows "N/A" in the workbook); any
mismatch is a transcription error and goes in the verification log.
- Write
scoring/findings-register.csv from the same validated leaf findings (one row
per finding: ID, severity, category, page, evidence, remediation, effort), per the shared
workspace contract. Neutralize leading =, +, -, @, tab and carriage-return characters
in text fields the same way (prefix ') so the CSV cannot run as a formula when opened.
- Hand off to
audit-reporting-wasp-drone (wave W8) once every applicable Wave-5/W6 Drone's
findings are either scored or logged as an unresolved rejection in the run's verification
log - never hand off with a silently-dropped finding.
References map
| Path |
Load when |
references/research/distilled-audit-scoring.md |
Verifying any formula-mechanics claim fast, or tracing where it came from |
references/research/raw/ |
Tracing a claim to its primary source |
guides/01-rollup-procedure.md |
Running or verifying the leaf-to-sub-audit-to-category-to-final rollup |
guides/02-critical-security-override.md |
Understanding or verifying the override mechanics, or retuning its two named ranges |
guides/03-nesting-structure-design.md |
Understanding which parts of the hierarchy are sourced vs. this Stinger's own design, before trusting or changing the structure |
guides/04-retuning-weights.md |
Retuning any category or sub-audit weight, the grade thresholds, or the override cap for one engagement |
guides/05-rejecting-a-leaf-finding.md |
Deciding whether an incoming leaf finding should be scored or rejected back to its originating Drone |
references/scripts/generate-scorecard-xlsx.py |
Regenerating the template after any design change; read its module docstring for the full sourced-vs-invented accounting |
references/scripts/README.md |
Full script inventory for this pair, including the shared plugin-root scripts this pair is listed against |
references/templates/website-audit-scorecard-template.xlsx |
The real, working branded XLSX template - copy this at the start of every run, never edit the template file itself |
references/templates/leaf-finding.schema.json |
The intermediate findings format every upstream Drone must write; validate every incoming finding against this before scoring it |
references/templates/README.md |
What each template file is for and how it is regenerated |
Related drones and stingers
- technical-seo-stinger - one of twelve Wave-5/W6 Drones whose leaf findings this pair consumes.
- aeo-audit-stinger - consumed, Wave 5.
- content-semantics-stinger - consumed, Wave 5.
- internal-linking-stinger - consumed, Wave 5.
- visual-funnel-stinger - consumed, Wave 5.
- accessibility-audit-stinger - consumed, Wave 5.
- web-security-posture-stinger - consumed, Wave 5; the sole source of the leaves that can trigger the critical-security-override.
- analytics-stack-stinger - consumed, Wave 5.
- performance-cwv-stinger - consumed, Wave 5.
- social-presence-stinger - consumed, Wave 5.
- blog-content-stinger - consumed, Wave 6a, conditional on blog detection.
- ecommerce-catalog-stinger - consumed, Wave 6b, conditional on commerce detection.
- audit-reporting-stinger - downstream consumer, Wave 8; renders this pair's populated
scoring/audit-scorecard.xlsx and findings-register.csv into the customer- and auditor-facing reports.
- audit-intake-stinger - upstream, Wave 0; scaffolds the shared workspace this pair reads from and writes into.
csv-xlsx-import-export-stinger (in the vibe-coding-tools plugin, not this one - no relative path resolves across plugins) - general-purpose CSV/XLSX mechanics; consult for openpyxl or spreadsheet-format questions beyond this pair's own scoring-specific research archive, per the build plan's "cite as related, reuse rather than duplicate" instruction for transferable knowledge the existing Wasp Nest roster already holds.
Critical Directive
- You must read all files and context contained within your skill.
- In the event your core knowledge does not provide sufficient guidance you must make every attempt to search the internet, related knowledge base documentation files, and other available resources to supplement your knowledge prior to proceeding with your task.
- Additional related skills can be found here:
Ship Gate decision
Does not apply in the usual "before committing code to this repo" sense: this pair's
day-to-day output (a populated scoring/audit-scorecard.xlsx inside an external customer's
audit workspace) is a client deliverable, not a change to this plugin's own source tree, so
the security-stinger / quality-stinger / github-repo-health-stinger Ship Gate sequence that
governs committing code has no natural trigger point in this pair's normal operation.
Stated explicitly rather than left implicit, because one part of this pair's own forge output
genuinely IS a committed repository artifact: the generated template file itself,
references/templates/website-audit-scorecard-template.xlsx, together with its generator
script, are real files committed to THIS plugin repository as part of Stage 4 (References).
Any future change to references/scripts/generate-scorecard-xlsx.py or to the weighting
design it encodes should go through this repo's normal commit discipline (and, if the change
touches code logic rather than pure data/content, the Ship Gate) before being committed -
this Stinger's own forge work in this session was reviewed for correctness by direct
verification (LibreOffice headless recalculation of the generated formulas, openpyxl
load-back, JSON Schema self-validation) rather than by the code-security/quality Ship Gate,
since it is template-generation tooling and reference content, not application source code
serving live traffic or handling secrets.
1---2name: audit-scoring-stinger-23description: "N/A-aware weighted rollup engine (leaf to sub-audit to category to final), critical-security-override, populates the branded named-range-driven XLSX scorecard."4license: AGPL-3.0-or-later5compatibility: "Claude Code, Cursor, ChatGPT Codex, Claude Cowork."6metadata:7 source: https://skillmd.com/skills/legioncodeinc/audit-scoring-stinger-28 hive-tier: stinger9 hive-drone: audit-scoring-wasp-drone10 research-window: "2026-08-18 (round 2 core-formula sweep; round 3 deeper sweep: N/A-aware formula mechanics, openpyxl branding mechanics, multi-level rollup design)"11 primary-surface: external-website-audit12---1314# Audit Scoring Stinger1516You are equipping **audit-scoring-wasp-drone**, part of the Website Auditor by Legion Code17Inc. plugin. Full scope and acceptance criteria: [prd-020-audit-scoring](../../library/requirements/backlog/prd-020-audit-scoring/prd-020-audit-scoring-index.md).1819This is one of the most load-bearing pairs in the plugin: wave W7, sync, needs every20applicable Wave-5/W6 Drone's findings before it can run at all, and its output (the populated21`scoring/audit-scorecard.xlsx`) is what `audit-reporting-wasp-drone` renders into both22customer- and auditor-facing deliverables in wave W8. Every factual claim below traces to a23downloaded primary source in `references/research/raw/`, or is explicitly flagged as this24Stinger's own engineering design where the research archive itself says the design is not25directly sourced - see `guides/03-nesting-structure-design.md` for the full accounting.2627## Purpose2829Roll every leaf finding (0-6 score, evidence pointer, one-line justification) up through30sub-audit, category, and final scores using N/A-aware masked-SUMPRODUCT formulas at every31level; apply the critical-security-override (any Security leaf scored 1 caps the final grade32at C); and populate the branded, named-range-driven XLSX scorecard, per build plan sections334.1-4.4 and prd-020's acceptance criteria.3435## When to use this skill3637- Wave W7 of every audit run, after every applicable Wave-5/W6 Drone has written its findings38 (`03-seo/` through `12-ecommerce/`, per the shared workspace contract).39- Verifying a final letter grade or the security-override cap by hand against the workbook's40 own formulas (e.g. during a QA pass on this pair, or when a client questions a grade).41- Retuning category or sub-audit weights via the `Rubric` sheet's named ranges for one42 engagement, without editing any formula (`guides/04-retuning-weights.md`).43- Validating an incoming leaf finding against `references/templates/leaf-finding.schema.json`44 and deciding whether to score it or reject it back to its originating Drone45 (`guides/05-rejecting-a-leaf-finding.md`).46- Regenerating the branded XLSX template itself after a design change to the category/47 sub-audit weighting structure (`references/scripts/generate-scorecard-xlsx.py`).4849## When not to use this skill5051- To re-score, second-guess, or reinterpret an upstream Drone's own leaf finding. That is the52 originating Drone's domain judgement, not this Stinger's arithmetic. A malformed or53 unevidenced finding is rejected back to its origin, never silently corrected here54 (`guides/05-rejecting-a-leaf-finding.md`).55- To generate the customer- or auditor-facing narrative reports. That is56 `audit-reporting-wasp-drone` / `audit-reporting-stinger`'s job in wave W8, reading this57 pair's output as its own input.58- To crawl, capture evidence, or produce a finding from scratch. This pair only consumes59 already-produced leaf findings; it has no domain expertise in security, SEO, accessibility,60 or any of the other audited domains.61- To determine category weights or the override rule from first principles. Both are binding62 product requirements straight from `plan/website-auditor-build-plan.md` section 4.2/4.3 and63 prd-020's acceptance criteria (AC-2, AC-3), not this Stinger's discretion to redesign; only64 the numeric VALUES in the named ranges are meant to be retuned per engagement.6566## Procedure67681. **Copy the template.** Copy `references/templates/website-audit-scorecard-template.xlsx`69 to the run's `scoring/audit-scorecard.xlsx` (build plan section 3's shared workspace path).70 Never edit the template in place; never hand-edit a formula cell in the copy either71 (`guides/04-retuning-weights.md` section 6).722. **Read every applicable category folder's output** (`03-seo/` through `12-ecommerce/`) and73 `_shared/evidence-index.md`, per the shared workspace contract in prd-020.743. **Validate each candidate leaf finding** against75 `references/templates/leaf-finding.schema.json`. A finding that fails validation, is a76 malformed boolean checkpoint, or targets an unknown category/sub-audit coordinate is77 rejected back to its `originating_bee` per `guides/05-rejecting-a-leaf-finding.md` - never78 scored anyway.794. **Transcribe every valid leaf finding** into its row on the `Scorecard` sheet (score,80 evidence pointer, justification - a direct write, never a formula edit). Evidence and81 justification text comes from third-party pages: write it as text, and prefix any value that82 starts with `=`, `+`, `-`, `@`, a tab or a carriage return with `'` so no cell becomes a formula.835. **Recalculate** (open in Excel/LibreOffice, or force a headless recalculation pass) and84 read back the rollups: sub-audit (`guides/01-rollup-procedure.md` section 1-2), category85 (section 3), and final (section 4). The critical-security-override86 (`guides/02-critical-security-override.md`) resolves automatically on the87 `Executive Scorecard` sheet from the same recalculation - no separate step is needed.886. **Verify, don't just trust.** Spot-check at least one sub-audit rollup and the final letter89 grade by hand against the raw leaf scores before treating the run as complete, per90 `guides/01-rollup-procedure.md` section 5's verification method. Cross-check the whole91 rollup with `shared/scripts/score-rollup.py` (see `references/scripts/README.md`). The92 script and the workbook apply the same N/A rule at every level, so they must agree93 exactly (an N/A sub-audit or category it lists shows "N/A" in the workbook); any94 mismatch is a transcription error and goes in the verification log.957. **Write `scoring/findings-register.csv`** from the same validated leaf findings (one row96 per finding: ID, severity, category, page, evidence, remediation, effort), per the shared97 workspace contract. Neutralize leading `=`, `+`, `-`, `@`, tab and carriage-return characters98 in text fields the same way (prefix `'`) so the CSV cannot run as a formula when opened.998. **Hand off to `audit-reporting-wasp-drone`** (wave W8) once every applicable Wave-5/W6 Drone's100 findings are either scored or logged as an unresolved rejection in the run's verification101 log - never hand off with a silently-dropped finding.102103## References map104105| Path | Load when |106|---|---|107| `references/research/distilled-audit-scoring.md` | Verifying any formula-mechanics claim fast, or tracing where it came from |108| `references/research/raw/` | Tracing a claim to its primary source |109| `guides/01-rollup-procedure.md` | Running or verifying the leaf-to-sub-audit-to-category-to-final rollup |110| `guides/02-critical-security-override.md` | Understanding or verifying the override mechanics, or retuning its two named ranges |111| `guides/03-nesting-structure-design.md` | Understanding which parts of the hierarchy are sourced vs. this Stinger's own design, before trusting or changing the structure |112| `guides/04-retuning-weights.md` | Retuning any category or sub-audit weight, the grade thresholds, or the override cap for one engagement |113| `guides/05-rejecting-a-leaf-finding.md` | Deciding whether an incoming leaf finding should be scored or rejected back to its originating Drone |114| `references/scripts/generate-scorecard-xlsx.py` | Regenerating the template after any design change; read its module docstring for the full sourced-vs-invented accounting |115| `references/scripts/README.md` | Full script inventory for this pair, including the shared plugin-root scripts this pair is listed against |116| `references/templates/website-audit-scorecard-template.xlsx` | The real, working branded XLSX template - copy this at the start of every run, never edit the template file itself |117| `references/templates/leaf-finding.schema.json` | The intermediate findings format every upstream Drone must write; validate every incoming finding against this before scoring it |118| `references/templates/README.md` | What each template file is for and how it is regenerated |119120## Related drones and stingers121122- [technical-seo-stinger](../technical-seo-stinger) - one of twelve Wave-5/W6 Drones whose leaf findings this pair consumes.123- [aeo-audit-stinger](../aeo-audit-stinger) - consumed, Wave 5.124- [content-semantics-stinger](../content-semantics-stinger) - consumed, Wave 5.125- [internal-linking-stinger](../internal-linking-stinger) - consumed, Wave 5.126- [visual-funnel-stinger](../visual-funnel-stinger) - consumed, Wave 5.127- [accessibility-audit-stinger](../accessibility-audit-stinger) - consumed, Wave 5.128- [web-security-posture-stinger](../web-security-posture-stinger) - consumed, Wave 5; the sole source of the leaves that can trigger the critical-security-override.129- [analytics-stack-stinger](../analytics-stack-stinger) - consumed, Wave 5.130- [performance-cwv-stinger](../performance-cwv-stinger) - consumed, Wave 5.131- [social-presence-stinger](../social-presence-stinger) - consumed, Wave 5.132- [blog-content-stinger](../blog-content-stinger) - consumed, Wave 6a, conditional on blog detection.133- [ecommerce-catalog-stinger](../ecommerce-catalog-stinger) - consumed, Wave 6b, conditional on commerce detection.134- [audit-reporting-stinger](../audit-reporting-stinger) - downstream consumer, Wave 8; renders this pair's populated `scoring/audit-scorecard.xlsx` and `findings-register.csv` into the customer- and auditor-facing reports.135- [audit-intake-stinger](../audit-intake-stinger) - upstream, Wave 0; scaffolds the shared workspace this pair reads from and writes into.136- `csv-xlsx-import-export-stinger` (in the `vibe-coding-tools` plugin, not this one - no relative path resolves across plugins) - general-purpose CSV/XLSX mechanics; consult for openpyxl or spreadsheet-format questions beyond this pair's own scoring-specific research archive, per the build plan's "cite as related, reuse rather than duplicate" instruction for transferable knowledge the existing Wasp Nest roster already holds.137138## Critical Directive139140- You must read all files and context contained within your skill.141- In the event your core knowledge does not provide sufficient guidance you must make every attempt to search the internet, related knowledge base documentation files, and other available resources to supplement your knowledge prior to proceeding with your task.142- Additional related skills can be found here:143 - [audit-scoring-wasp-drone](../../agents/audit-scoring-wasp-drone.md) - this Stinger's paired Drone.144145## Ship Gate decision146147Does not apply in the usual "before committing code to this repo" sense: this pair's148day-to-day output (a populated `scoring/audit-scorecard.xlsx` inside an external customer's149audit workspace) is a client deliverable, not a change to this plugin's own source tree, so150the security-stinger / quality-stinger / github-repo-health-stinger Ship Gate sequence that151governs committing code has no natural trigger point in this pair's normal operation.152153**Stated explicitly rather than left implicit, because one part of this pair's own forge output154genuinely IS a committed repository artifact:** the generated template file itself,155`references/templates/website-audit-scorecard-template.xlsx`, together with its generator156script, are real files committed to THIS plugin repository as part of Stage 4 (References).157Any future change to `references/scripts/generate-scorecard-xlsx.py` or to the weighting158design it encodes should go through this repo's normal commit discipline (and, if the change159touches code logic rather than pure data/content, the Ship Gate) before being committed -160this Stinger's own forge work in this session was reviewed for correctness by direct161verification (LibreOffice headless recalculation of the generated formulas, `openpyxl`162load-back, JSON Schema self-validation) rather than by the code-security/quality Ship Gate,163since it is template-generation tooling and reference content, not application source code164serving live traffic or handling secrets.