Gap Analysis

Gap Analysis

This skill produces two related deliverables that live in the project's gap-analysis folder (referred to as `` throughout — the actual path is project-specific and supplied by the user; see "Required inputs" below):

1. Individual gap-analysis reports — one HTML file per workflow / component / system, comparing a BPMN (or set of requirement docs) to the RFP and Detailed Requirements, listing gaps with severity, workstream, and remediation owner.
2. The master gap & risk register (gap-register.html) — a roll-up index that re-tabulates every gap from every individual report, grouped by category, with a scorecard and filter bar.

Both deliverables share the same theme system, severity rubric, workstream pills, and owner pills. Pick the right flow based on the user's intent:

User asks for a fresh gap analysis on a specific workflow / component
+ no existing report at /-gap-analysis.html
        │
        ▼
   Flow A — Fresh individual report
        │
        ├──→ assets/individual-report.template.html
        └──→ references/individual-report.md

User asks to update / add to / refresh an existing individual report
(usually points at /-gap-analysis.html)
        │
        ▼
   Flow A2 — Update existing report
        │
        └──→ references/individual-report.md (§ Updating an existing report)

User points at gap-analysis folder, OR mentions "register" / "index"
+ wants the master rollup refreshed
        │
        ▼
   Flow B — Master register
        │
        ├──→ assets/register.template.html
        └──→ references/register.md

If the user asks to "do gap analysis" without specifying which, ask once before generating. If the user names a specific topic, check whether /-gap-analysis.html already exists — if it does, default to Flow A2 (update) rather than Flow A (fresh).

---

Required inputs

Two things are project-specific and cannot be embedded — they must come from the user at invocation time. Ask once at the start of any flow for whichever is missing; confirm the path explicitly rather than guessing or inferring it from a filename in the prompt.

1. **`** — where reports live. This document uses the token throughout; substitute the path the user gives. Engagements lay out output/ differently (output/requirement/gap-analysis/, output/gap-analysis/, docs/gap-analysis/, …), so a wrong path silently writes into the wrong place or fails to find existing reports on a register refresh. Ask: "What's the project's gap-analysis output folder — the one that holds (or will hold) the *-gap-analysis.html files and gap-register.html`?"

1. Glossary / ubiquitous-language file — the terminology source of truth. Every engagement uses different terms, acronyms, and cross-document synonyms, so a hardcoded path produces wrong analysis on a new project. It may live anywhere (output/requirement/ubiquitous-language.md, docs/glossary.md, embedded in a README, …). Ask: "Where's the glossary / ubiquitous-language file — the one listing canonical terms and known synonyms?" Read it in full before drafting any row whose verdict turns on a term (including every AMB- decision). conventions.md §6 is the source of truth for glossary discipline.

If the user genuinely has no glossary: proceed, default every terminology mismatch to AMB-, and note up-front in the report intro — "No glossary supplied — terminology variants flagged conservatively; reconcile against a glossary before circulating widely."

---

Shared conventions

The reference file references/conventions.md is the source of truth for severity rubric, gap-ID prefixes, pill class forms, owner pills, card types, glossary discipline, provenance rules, and the row ↔ arch-box coupling rule. Read the sections relevant to your flow — don't read what you won't use:

| Flow | Sections to read first | |------|------------------------| | A — Fresh report | §1 severity · §2 prefixes · §3 pills · §4 owners · §5 cards · §6 glossary · §7 provenance · §8 coupling | | A2 — Update existing report | §2 prefixes (decision tree only) · §3 pills (table + filename rule) · §8 coupling. Skip §5 cards unless you're adding a new phase-flow card. | | B — Master register | §1 severity · §2 prefixes · §3 pills · §4 owners · §6 glossary · §7 provenance |

The legacy R- IDs preserved in some earlier reports are never renumbered; new SAP-integration payload blockers use SAP-.

Two cross-cutting rules from this workspace's CLAUDE.md apply to all flows, but the form differs by deliverable:

• Provenance. In HTML reports (this skill), provenance lives inline in the gap row itself: the FR / RFP / BPMN citation goes in the Gap column wrapped in ..., and the Why It Matters sentence frames why the citation matters for the bid. Do not insert Markdown blockquotes (> **RFP source reference:** ...) into HTML — those are the convention for the workspace's Markdown deliverables (e.g. rfp-2026.md) and look out-of-place in the gap-analysis house style.
• Inconsistencies. Surface contradictions between source documents rather than papering over them. Where documents disagree on a term, preserve both variants verbatim and tag the row AMB- rather than silently picking one — see conventions.md §6 for the full glossary discipline.

---

Flow A — Fresh individual gap analysis report

Use when the user asks to gap-analyse a specific workflow, component, sub-system, or BPMN file and no report already exists for that topic. Examples:

• "Analyse WS5-01_Quote Creation.bpmn against the detailed requirements and create a gap analysis report."
• "Compare RFP.md and Detailed_Requirements_FINAL.md for the standard approval workflow and produce a gap analysis."

If a report on this topic already exists in `` (e.g. the user is asking to re-run on a topic that's been done before), switch to Flow A2 — preserving existing gap IDs and the historical record matters more than starting clean.

Steps

1. Locate inputs. Read every requirement / BPMN / FeatureSpec / RFP document the user named, plus the glossary file the user supplied (see "Required inputs" above; ask if absent). Convert binaries (.docx, .xlsx) per the project's conversion conventions before reading.
2. Pick a slug for the filename: -gap-analysis.html, kebab-case, no spaces. New file lives in `` (the path the user supplied — see "Required inputs" above; ask if absent).
3. Copy assets/individual-report.template.html to the output path. The template is self-contained — theme CSS, theme toggle, and the JS block are ready as-is. Do not strip them.
4. Customise the title, subtitle (source files + date), legend, phase rows, gap-register table, architecture summary, and reading notes. See references/individual-report.md for the full section anatomy and what each card type / badge means.
5. Mirror every row to the arch boxes. Per conventions.md §8: every row in the gap-register table also needs a matching bullet in box-gap / box-implied / box-confirmed. Don't save until both views are in sync.
6. Re-roll up. After saving, offer to run Flow B to refresh gap-register.html so the master index reflects the new findings.

---

Flow A2 — Update existing individual report

Use when the user names a specific existing report and asks to add a finding, change severity, mark a gap resolved, or refresh the date. Examples:

• "Update the approval-workflow gap analysis with one new High-severity row about JARVIS error-handling."
• "Mark G-04 in the quote-creation gap analysis as resolved — the FX-rate question was answered."
• "Refresh the rebate-accounting report date to today."

This flow is deliberately narrow. Do not re-read the full RFP / BPMN source unless the new finding asserts a fact you cannot verify from the user's prompt or the existing report. Treat the existing report as authoritative for everything else.

Steps

1. Read the existing HTML report only. Do not pull source documents unless the user's instruction names a specific FR/RFP/BPMN reference you have not seen before and need to verify. If the new finding hinges on terminology (you're flagging an AMB- ambiguity, or you're not sure whether two terms are synonyms), also read the glossary file the user supplied.
2. Locate the change site. Find the Gap Register ` and the arch-row containing box-confirmed / box-implied / box-gap`. These two regions are the entire change surface for most updates.
3. For a new gap row:

• Pick the prefix using the decision tree in conventions.md §2 (BPMN modelling defect → BPM-, missing FR → MFR-, two-doc contradiction → CTR-, etc. — G- is the default, not the first choice).
• Pick the next available number for that prefix by scanning the existing rows; never reuse or renumber.
• Place the row after the last existing row of the same severity (Critical → High → Medium → Low), then by ascending ID within severity.
• Use the individual-report pill form: WS4, not the master-register ws ws-ws4 form. See conventions.md §3.
• Severity uses the bare-class cell form here: High — not the badge form (which is reserved for the master register).

1. Mirror the row to the matching arch box (per conventions.md §8). Open gaps → box-gap. Inconsistencies → box-implied. Newly verified items → box-confirmed. Bullet format: Short title (BPM-04). Skipping this step is the single most common reason updates fail review.
2. For severity changes on an existing row: update both the row's severity class and move the matching arch-box bullet if the box-mapping changed (e.g. resolving an open gap → bullet moves from box-gap to box-confirmed).
3. For "mark resolved": change the row to Resolved and add class="resolved" to the ``. Do not delete the row — the historical record matters. Update the arch box per step 4.
4. Update the subtitle date to today (the user usually says so, but do it even when they don't — the date is the audit trail).
5. Save and stop. Do not re-write unrelated sections, do not "improve" surrounding prose, do not refactor the legend. Surgical changes only.

After Flow A2, offer to run Flow B if the change was material (new Critical / High row, new prefix family introduced).

---

Flow B — Master gap & risk register

Use when the user points at the gap-analysis folder, says "regenerate the index", "refresh the register", or otherwise asks for the cross-file rollup.

Steps

1. Discover inputs. List /*.html (excluding gap-register.html itself), where `` is the path the user supplied — see "Required inputs" above; ask if absent. Read each report and extract every row from its gap-register table.
2. Re-categorise. The register is grouped by category (e.g. "SAP Integration — Payload Design Blockers", "WS6 Rebate Accounting", "Workflow Configuration Risk", etc.) — not by source file. Map each gap into the existing category set first; only create a new category when the existing seven do not fit. See references/register.md for the canonical category list and category numbering colours.
3. Promote IDs as needed. When a gap appears in multiple individual reports, keep one canonical ID and note the cross-references. Where gaps in different reports describe the same underlying issue, dedupe.
4. Recompute the scorecard. Count gaps by severity (Critical / High / Medium / Open Decision / Resolved+Low) and list IDs in each tile.
5. Add per-report quick links. The template already includes a ` block (after the scorecard, before the filter bar) and the matching .report-links CSS rule. Replace its example tags with one -gap-analysis.html"> per individual report discovered in step 1 (exclude gap-register.html` itself). Link text is the human-readable topic name (e.g. "Quote Creation", "Approval Workflow", "SAP Integration"). Keep this list in the same order as the categories below where practical.
6. Copy assets/register.template.html to /gap-register.html and fill in scorecard, report links (step 5), filter bar, category headers, and category tables. The filter JS is pre-wired — keep the data-sev and data-ws attributes consistent with the filter buttons.
7. Footer pointer. Update the footer's last-updated date and any pointers to companion reports.

---

What "good" looks like

Both deliverables are bid-support artefacts — clarity for BSC and the bid team is more important than completeness. A good report:

• Names the contradiction explicitly ("RFP §5 says 1,200 concurrent users; NFR-P02 says 200 — 6× discrepancy") rather than picking a value.
• Ties every gap to a source citation (FR ID, NFR ID, RFP section, BPMN task ID). A gap with no source pointer is a gap the reader cannot verify.
• Distinguishes blocker from concern in the Why It Matters column. "Vendor cannot start build until BSC answers this" is different from "audit trail is weaker than ideal".
• Surfaces missing FRs separately from contradicted FRs (e.g. MFR- prefix for "FeatureSpec P0, no FR exists"). The bid team needs both lists for different reasons.
• Keeps the open-question loop closed. Every Critical / High gap should have a counterpart in the project's open-questions file (typically a sibling of `, e.g. /open-questions.md`, grouped A–G). Mention this to the user when finishing — and ask for the open-questions file path if you don't already have it.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: kevinlin
• Source: kevinlin/skills
• License: MIT

Install and usage instructions live in the source repository linked above.