Forty-five invocable skills. Shared packs under `core/skills/_shared/` are not skills. The architect, database, security, repo-analyst, and shell-runner files under `core/agents/` are roster roles spawned by O1 or the router, not slash skills. The long form of a skill is the session linked on its heading. This page is the index that still says what the skill does, what it writes, and which skill it calls next.
Installs into .claude/skills of the current project.
Are you the author of Sessions?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/tibursocampos-sessions)
# 09 — Every skill
Forty-five invocable skills. Shared packs under `core/skills/_shared/` are not skills. The architect, database, security, repo-analyst, and shell-runner files under `core/agents/` are roster roles spawned by O1 or the router, not slash skills.
The long form of a skill is the session linked on its heading. This page is the index that still says what the skill does, what it writes, and which skill it calls next.
## Delivery
### `memory-bank-init`
Creates or refreshes `memory-bank/` at the repository root or the global classic path. Never under `features/NNN-slug/`. Modes: `create`, `refresh`, `refresh-light`. Runs a read-only inventory, then fills templates. Confirms before write. Stops when the bank is done and does not start O1. Called at Orchestrated Delivery Step 0 and after O3 changes application files (`refresh-light`). Detail: [01](01-orchestrated-delivery.md).
### `orchestrate-analyze`
O1. Classifies intent, asks a few high-cost questions, sets nature and `needs_*`, spawns specialists only when flags require them, and holds their notes until the feature has no open question. Then it creates story folders, drafts architecture for greenfield and waits for **sim**, merges stories with scorecard and promotion gates, and waits for backlog **sim**. Writes `FEATURE.md`, `CONTINUITY.md`, and `STORY.md`. Does not write PRD, PLAN, or code. Detail: [01](01-orchestrated-delivery.md).
### `orchestrate-deliver`
O2. Requires an approved backlog and a feature with no open question. For each story, checks the story files, runs `sdd-spec`, contests that PRD, then runs `sdd-plan`. An open question, including **MINOR**, stops that story. Series writes in the parent after **sim**. Parallel children return drafts one gate at a time; the parent is the only writer. Stops on missing specialist folders. Brownfield also writes `CHANGE.md`. Preflight must pass before O3. Detail: [01](01-orchestrated-delivery.md), contracts in [02](02-classic-sdd.md).
### `orchestrate-develop`
O3. Builds a queue of `PENDING` PLAN steps. Before each step, calls `repo-analyst` and `architect`, and `database` when the plan cites persistence. Contract drift stops with `plan_stale` and returns to `sdd-plan`. Spawns one `sdd-develop` child per step. Parent writes no application code. Execution mode `serial` (default), `parallel`, or `manual`. Pacing `step_by_step` (default) or `continuous`. After code changes, offers memory-bank refresh-light. When the feature is done, the order is `code-review`, `run-tests`, `security`, then `/commit` and `/push`. Detail: [01](01-orchestrated-delivery.md).
### `sdd-spec`
Writes one PRD under `features/NNN-slug/.../PRD/`. What and why, stable `REQ-NNN`, out of scope, confirm before write, `validate-prd`. Brownfield also writes `CHANGE.md`. Does not write a PLAN in this session. Direct use can promote a cited markdown file without O1. Orchestrated use returns to O1 when specialist folders are missing. Detail: [02](02-classic-sdd.md).
### `sdd-plan`
Writes one PLAN beside that PRD. Copies step ids from `split-story-checklist`. Every REQ maps to one of those steps. Two reviews before **sim**. Bodies of SQL and OpenAPI stay in canonical files; the PLAN cites them. `validate-plan` must pass. Does not implement and does not record duration. Detail: [02](02-classic-sdd.md).
### `sdd-develop`
Implements exactly one PLAN step, updates that PLAN, runs targeted tests. Optional evidence files and TRACE when the wave closes. Does not start the next step. This is the contract O3 children follow. Detail: [02](02-classic-sdd.md).
### `read-sdd-artifact`
Read-only. Turns one FEATURE, STORY, PRD, or PLAN path under `features/` into `source_context`. Rejects traversal, absolute paths, and other kinds. No confirm gate. Consumers that already have the envelope skip an opaque re-read. Detail: [02](02-classic-sdd.md).
### `refine-story`
Standalone shaping of one bug, user story, or technical story. On a direct invoke, mode `feature`, `tech`, or `split` is asked in the chat language. An orchestrated call receives `mode=feature` or `mode=tech` and does not ask. Writes chat markdown, a scorecard, and optionally `STORY.md` or `docs/backlog/<slug>.md`. O1 uses the scorecard rubric file without calling this skill. O2 calls this skill when a question is still open. Detail: [03](03-backlog-shape.md).
### `split-story-checklist`
Turns a story into a dependency checklist (`REFINE/tasks.md`). When `sdd-plan` calls it with `source=prd`, groups come from the PRD even if no steps exist yet, and a trivial story still returns one step. A direct invoke with no steps returns to `/refine-story`. Does not create new story folders. At most five groups. Asks documentation language only on a direct invoke. Detail: [03](03-backlog-shape.md).
## Implementation
### `developer`
Router. UI brief check, then the stack table, then ad-hoc fallback for scripts and HTML. Large scope goes to `sdd-spec`. Trivial O1 triage may land here. It does not implement PLAN steps. Detail: [04](04-implement-and-guidelines.md).
### `dotnet-developer`
Small or medium .NET without a full PLAN. Loads one `dotnet-guidelines` style file. Build and tests. Scope growth returns to `sdd-spec`. Also the fix path when `test-coverage` fails on C#. Detail: [04](04-implement-and-guidelines.md).
### `java-developer`
Small or medium Java. Spring Boot is the default, not the only shape the skill names. Loads `java-guidelines` including one architecture overlay. Detail: [04](04-implement-and-guidelines.md).
### `javascript-developer`
Small or medium Node or DOM work. There is no `node-developer` id. Loads `javascript-guidelines`. Nest is recognized only as an existing project signal, not as a separate skill. Detail: [04](04-implement-and-guidelines.md).
### `python-developer`
Small or medium Python (FastAPI or Flask, pytest). Loads `python-guidelines` and one shared Layer B style file. Detail: [04](04-implement-and-guidelines.md).
### `react-developer`
Small or medium React web (hooks, tests). Reads `DESIGN-BRIEF.md` when present. Existing Blip plugins (`blip-ds`) also load `blip-guidelines`. Mobile work is `react-native-developer`. Detail: [04](04-implement-and-guidelines.md).
### `react-native-developer`
Small or medium React Native or Expo. Uses the React Native guideline pack, not the web React pack, and does not load Blip guidelines. Detail: [04](04-implement-and-guidelines.md).
### `angular-developer`
Small or medium Angular (standalone components, signals, RxJS). Loads `angular-guidelines` and HTML/CSS files as needed. Detail: [04](04-implement-and-guidelines.md).
### `vue-developer`
Small or medium Vue 3 (Composition API, Pinia, Vitest). Detail: [04](04-implement-and-guidelines.md).
### `blazor-developer`
Small or medium Blazor UI (WASM, Server, Hybrid). Backend API work in the same change hands to `dotnet-developer`. Detail: [04](04-implement-and-guidelines.md).
### `electron-developer`
Small or medium Electron (main, preload, renderer, IPC, packaging). Security and CSP load first when IPC changes. One renderer pack (React, Vue, or JS) is loaded, not all of them. Detail: [04](04-implement-and-guidelines.md).
### `arduino-developer`
Configuration-aware Arduino sketches, libraries, and Arduino Core firmware. Requires evidence for the target, core/version, tooling, provenance, and validation boundary; native ESP-IDF and MicroPython work hand off to their dedicated skills. Detail: [04](04-implement-and-guidelines.md).
### `esp-idf-developer`
Native ESP-IDF C/C++ firmware using the project’s declared target, version, CMake configuration, `idf.py` workflow, and Component Manager metadata. Keeps host, compile, device, and HIL evidence separate. Detail: [04](04-implement-and-guidelines.md).
### `micropython-developer`
MicroPython firmware for identified ESP32 or ESP8266 targets. Keeps host CPython tooling/tests, deployment, and device-runtime evidence separate and does not generalize APIs across ports or builds. Detail: [04](04-implement-and-guidelines.md).
### `blip-plugin-developer`
Scaffolds a **new** Blip React plugin and records the profile. Does not implement the feature set. Hands UI to `react-developer`, a new screen’s brief to `impeccable`, a backend to `dotnet-developer`, and a full spec to `sdd-spec`. An existing `blip-ds` project starts at `react-developer` instead. Detail: [04](04-implement-and-guidelines.md).
### `impeccable`
Design commands (`init`, `shape`, `craft`, `critique`, `audit`, `harden`, and others). Writes `docs/DESIGN-BRIEF.md` after confirm and stops. The next chat is the stack skill named by `target_stack`. Does not implement in the same session. Detail: [04](04-implement-and-guidelines.md).
## Review and change
### `code-review`
Review of a branch against guidelines and, when found, PRD/PLAN. Asks single versus multi-angle. Findings name a file and a line, severity `critical`, `important`, or `advisory`. A `critical` finding stays blocking until a fix and a new review, at most three rounds. Decisions: Approved, Approved with reservations, Changes required. Does not edit code. First step when an O3 scope closes. Detail: [05](05-review-and-quality.md).
### `test-coverage`
.NET Coverlet report. Default 80% line coverage on changed production files. Writes `TestResults/CoverageReport/`. Does not block merge by itself; `code-review` applies the threshold. Called from `run-tests` only when the PLAN asks for coverage. Detail: [05](05-review-and-quality.md).
### `run-tests`
Runs the test command of each detected stack and returns `PASS` or `FAIL`. Does not edit code and does not replace `test-coverage`. Used at the end of an O3 scope, after `code-review`. Detail: [05](05-review-and-quality.md).
### `repair-dotnet-build`
Local `dotnet build` / `dotnet test`, or a pasted log. One error at a time. Proposes each fix and waits. Does not call a remote CI API and does not cover Node or Python. Detail: [05](05-review-and-quality.md).
### `refactor`
One safe refactor step with tests still green. Does not mix in features or bugfixes. No auto-commit. Detail: [05](05-review-and-quality.md).
### `performance-profile`
Finds a hot path, waits for a workflow choice, then proves the change with a micro-benchmark. Does not bypass a domain rule for speed. Detail: [05](05-review-and-quality.md).
### `framework-upgrade`
Modes `audit`, `plan`, `migrate`, `validate`. Packs on disk: `angular`, `dotnet`. Migrate needs **sim**. The skill id never includes a version number. Detail: [06](06-framework-upgrade.md).
## Platform
### `api-standards`
Agnostic HTTP design (shape, versioning, errors, naming, security hygiene). Does not generate clients and does not embed a company contract. Hands generation to `api-integrate`. Detail: [07](07-platform-skills.md).
### `api-integrate`
Typed client and DTOs from OpenAPI or Swagger. Waits for a workflow choice. No secrets in source. Standards questions return to `api-standards`. Detail: [07](07-platform-skills.md).
### `i18n-manager`
Extracts UI string literals into `.resx` or `.json` and replaces them with translation keys. Waits for a workflow choice before writing. Does not localize logs or configuration. Detail: [07](07-platform-skills.md).
### `containerize`
Multi-stage Dockerfile, `.dockerignore`, and compose for local dependencies. Waits for a workflow choice. No secrets in files that will be committed. Detail: [07](07-platform-skills.md).
### `ef-add-migration`
Discovers startup project, DbContext, and migrations folder, then `dotnet ef migrations add`. Confirms an inferred name. Consumer repo, not this toolkit, unless the toolkit is the subject. Detail: [07](07-platform-skills.md).
### `scaffold-message-handler`
Collects queue, contract, retry, and idempotency requirements before writing a consumer. Detects MassTransit, RabbitMQ, or Service Bus from the consumer repo. Does not ship a corporate template. Detail: [07](07-platform-skills.md).
## Git, catalog, docs
### `help-skills`
Prints the static catalog (`CATALOG.md`, and `OPERATOR.md` when the question is about confirmations). Does not invent ids and does not load every skill body. Detail: [08](08-git-and-docs.md).
### `commit`
Conventional Commit on a valid feature branch after the message is approved. Asks about memory-bank and project docs when those trees exist. Strips any `Co-authored-by` trailer. Does not open a pull request. Detail: [08](08-git-and-docs.md).
### `push`
`git push -u origin HEAD` after the branch check. If a pull request was already requested, enters `open-github-pr`. Otherwise asks. Detail: [08](08-git-and-docs.md).
### `open-github-pr`
Creates the pull request with `gh`. Feature into `develop` uses squash when auto-merge is enabled. Release from `develop` into `master` or `main` uses rebase. Asks for content **sim** and, separately, whether to enable auto-merge. Detail: [08](08-git-and-docs.md).
### `document-plan`
Asks documentation language once, then writes `docs/overview.md` and `docs/documentation-plan/plan.md`. Does not write feature PLANs. For this repository the language is English. Detail: [08](08-git-and-docs.md).
### `document-implement`
Executes one pending step of that documentation plan, then stops. Detail: [08](08-git-and-docs.md).
## Count
`CATALOG.md` groups the same 45 ids: Classic SDD 4, Backlog Refine 2, Orchestrated Delivery 4, developer routing and stack 14, Blip and design 2, operational 17, documentation 2.