inventory
/inventory
Skanuje bazę kodu trzykrotnie i tworzy jeden raport HTML z siedmioma zakładkami: architektura, moduły, schemat bazy, autorzy commitów, dług techniczny, obserwowalność i testy.
- Autor
- diegopacheco ↗
- Repozytorium
- ai-playground ↗
- Licencja
- Unlicense
- Ostatnia zmiana repozytorium
- 23.10.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 4
- Skrypty
- 2
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 23.10.2026, bez odwołań do wycofanych modeli i API.
Unlicense, otwarta.
- 2 skrypty w paczce. Przejrzyj je przed instalacją, bo Claude może je uruchomić.
Skrypty: scripts/render.py, scripts/scan.py
Co robi i kiedy
Skanuje bazę kodu trzykrotnie i tworzy jeden raport HTML z siedmioma zakładkami: architektura, moduły, schemat bazy, autorzy commitów, dług techniczny, obserwowalność i testy.
Oryginalny opis (EN)
Scans a codebase three times and produces one self-contained light-theme searchable HTML report with seven tabs - architecture overview with a hand-drawn diagram, core modules with pros and cons, database schema, committers, tech debt, observability and tests. Use when the user runs /inventory or asks for a codebase inventory, an architecture overview, a tech-debt map, a module breakdown, or a full audit of what a repository contains.
Jak działa
- Step 0 — set up
- Step 1 — pass 1, collect
- Step 2 — pass 2, verify
- Step 3 — pass 3, adversarial
- Step 4 — render
- Step 5 — hand it over
- What goes in each tab
- Rules
Podgląd SKILL.md
---
name: inventory
description: Scans a codebase three times and produces one self-contained light-theme searchable HTML report with seven tabs - architecture overview with a hand-drawn diagram, core modules with pros and cons, database schema, committers, tech debt, observability and tests. Use when the user runs /inventory or asks for a codebase inventory, an architecture overview, a tech-debt map, a module breakdown, or a full audit of what a repository contains.
allowed-tools: [Bash, Read, Write, Edit, Glob, Grep]
---
# Codebase Inventory
Three scan passes over a codebase, then one HTML report. The passes are not
optional and not merged: each has a different job and the third one throws
work away.
Argument: an optional path to scan. Default is the current directory.
## Step 0 — set up
```bash
SKILL_DIR="$(dirname "$0")" # the directory holding this SKILL.md
TARGET="$(cd "${1:-.}" && pwd)"
REPORT_DIR="$(mktemp -d "${TMPDIR:-/tmp}/inventory-$(basename "$TARGET")-XXXXXX")"
python3 "$SKILL_DIR/scripts/scan.py" "$TARGET" "$REPORT_DIR/facts.json"
```
Resolve `SKILL_DIR` for real; it is where `scripts/`, `assets/` and `prompts/`
live. Print `REPORT_DIR` to the user now — it is where the report will land and
it is also printed at the top of the report itself.
## Step 1 — pass 1, collect
Read `prompts/pass-1-collect.md` and follow it. Substitute
`$facts_json` = `$REPORT_DIR/facts.json`, `$repo_root` = `$TARGET`,
`$report_dir` = `$REPORT_DIR`. The output contract is `prompts/schema.md`.
Read real files. `facts.json` tells you where to look; it does not tell you
what the code means.
## Step 2 — pass 2, verify
Read `prompts/pass-2-verify.md` and follow it. Every path is checked against
disk, every number against `facts.json`. Corrections are recorded in
`verification.notes`.
## Step 3 — pass 3, adversarial
Read `prompts/pass-3-adversarial.md` and follow it. Cut anything unproven,
resolve contradictions between tabs, enforce the writing rules, rank and cap.
Set `verification.passes` to 3.
## Step 4 — render
```bash
python3 "$SKILL_DIR/scripts/render.py" \
"$REPORT_DIR/facts.json" "$REPORT_DIR/analysis.json" "$REPORT_DIR"
```
The renderer validates before it writes. When it exits non-zero it prints the
exact problems: a path that does not exist, a module without 5 pros and 5 cons,
a count that disagrees with the scan, an edge to a node that is not there. Fix
`analysis.json` and run it again.
Never edit `render.py` or `template.html` to make a report pass. The validator
is the point.
## Step 5 — hand it over
```bash
open "$REPORT_DIR/index.html" # macOS
```
Tell the user the report directory path and give a short summary: module count,
the top three tech-debt items, the test picture, and the one thing you would
fix first.
## What goes in each tab
1. **Architecture** — hand-drawn diagram from `architecture.nodes` and `.edges`,
plus the evidence table. Nodes are real things in the repo. Layout is
generated, so keep node labels short and set `layer` when you know the flow.
2. **Modules** — a card per core module with an emoji icon; the modal holds the
3-5 line description, the main files with full paths, 5 pros and 5 cons
where each con carries a 2-line reason and a 3-line fix.
3. **Schema** — only when a schema was found. Table count, main tables, size,
worst queries with the reason and fix, 5 pros and 5 cons.
4. **Committers** — avatar and name, ordered by commit count, with what each
person works on. Everything clickable.
5. **Tech debt** — up to 10 items per module, searchable cards; the modal holds
why it is bad, how to fix it, the anti-patterns behind it, and real examples
with file path and line.
6. **Observability** — a score for logging and a score for defensive code,
with findings, plus the dashboards and alerts found in the repository.
7. **Tests** — files and cases per test type, what is covered as x of y, and
the top 10 problems with the test suite itself.
## Rules
* No metaphors anywhere in the report. Say the mechanism.
* Every path in the report must exist. The validator enforces this.
* Never invent a number. Counts come from `facts.json`.
* When a section has no data, say so; do not fill it with generic advice.
* The report is one file. Do not add assets next to it.
Pliki w paczce (4)
- SKILL.md 4 KB
- assets/template.html 33 KB
- scripts/render.py 17 KB
- scripts/scan.py 25 KB
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · diegopacheco/ai-playground | 100 | Unlicense |
| ericrisco/rsc-harness | 100 | MIT |
| koda-ai-studio/agent-bom | 100 | Apache-2.0 |
| neurawork-git/n8n-autopilot | 100 | MIT |
| surenode-ai/skill-discovery | 100 | Apache-2.0 |
| koda-ai-studio/agent-bom | 86 | Apache-2.0 |
| koda-ai-studio/agent-bom | 86 | Apache-2.0 |
| melodic-software/claude-code-plugins | 84 | MIT |
Migawka bazy z 13.11.2026 · wersja paczki 7a52dcb030d2 (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100tech-debt-discovery
Inwentaryzuje i priorytetyzuje dług techniczny: skanuje znaczniki w kodzie, analizuje historię git, sprawdza zależności i tworzy priorytetowy raport hotspotów. - 100team-tech-debt
Wykrywa i naprawia dług techniczny zespołem agentów: skanuje kod, ocenia dotkliwość, planuje i wykonuje poprawki z walidacją, po poleceniu „team tech debt”. - 100team-tech-debt
Wykrywa i naprawia dług techniczny zespołem agentów: skanuje kod, ocenia dotkliwość, planuje i wykonuje poprawki z walidacją, po poleceniu „team tech debt”. - 100technical-debt
Inwentaryzuje i priorytetyzuje dług techniczny w projektach PHP/Laravel i Node/TypeScript/React, z audytem bezpieczeństwa, wydajności i danych. - 100manage-tech-debt
Śledzi, kategoryzuje i priorytetyzuje dług techniczny, tworzy rejestr długu, ocenia jakość kodu i planuje prace refaktoryzacyjne. - 100doncheli-tech-debt
Identyfikuje, ilościowo ocenia i priorytetyzuje dług techniczny w bazie kodu, w tym zapachy kodu i kandydatów do refaktoryzacji.