Skip to content
Back to skills

Devrites Lib

ASecurity

Internal shared DevRites helper library. Documents cross-cutting engine commands and references; not a user workflow. Do not invoke directly.

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 24, 2026
developmentrustshellfrontendsecurity

Works with

  • cursor

Security analysis

A100/100

Pro scans all 20 files and shows the line behind each finding

Scanned November 11, 2026

npx -y skills add ViktorsBaikers/DevRites --skill devrites-lib --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Devrites Lib?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Devrites Lib
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/viktorsbaikers-devrites-lib-devrites/badge)](https://www.skillsdirectory.com/skills/viktorsbaikers-devrites-lib-devrites)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: devrites-lib
description: Internal shared DevRites helper library. Documents cross-cutting engine commands and references; not a user workflow. Do not invoke directly.
triggers: []
---

# devrites-lib: shared workflow contracts

Do **not** invoke this skill directly.

## Ownership boundary

- The host owns instruction loading, exact-agent dispatch, scheduling, waiting,
  follow-up, results, and history. Skills state the role/result, never native
  fields or dispatch receipts.
- Dispatch every workflow-required role fresh. A missing role stops for HITL;
  never skip, substitute, or perform its work in the root.
- The root owns DevRites state/artifacts but never source/tests. Codex grants it
  workspace permission only because children cannot elevate; writing follows
  [`reference/standards/agents.md`](reference/standards/agents.md).
- The engine owns the retained deterministic structure, atomic-write, install,
  evidence-freshness, and safety primitives only.

## Workspace orientation

Use the supplied slug or read `.devrites/ACTIVE`. Require its authoritative
`state.md`; read it directly, then only needed phase artifacts. Never infer
lifecycle state from chat or optional `README.md`.

## Shared references

- [`reference/standards/agents.md`](reference/standards/agents.md): native custom-agent roles,
  immutable inputs, result contracts, and source-boundary review.
- [`reference/candidate-integrity.md`](reference/candidate-integrity.md): the
  content-bound candidate lifecycle from Build through Ship
  ([`workspace-artifact-schema.md`](reference/workspace-artifact-schema.md)).
- [`reference/reply-contract.md`](reference/reply-contract.md): compact user-facing
  completion states. The host renders the response normally.
- [`reference/parallel-workspaces.md`](reference/parallel-workspaces.md): explicit
  phase targets, one isolated global controller, and source-owned proof.
- [Autocomplete loop](../rite-autocomplete/reference/loop.md#early-preparation-admission):
  controller-only early preparation, bounded successor routing and normalization research.
- [`reference/visual-playbooks/index.md`](reference/visual-playbooks/index.md): progressive
  visual HTML playbook router (load matching ids only; dual-read outline).

## Deterministic engine surface

The engine is limited to:

- Checks: `check candidate` (content-bound identity), `check readiness`
  (structure; `--emit-binding` emits the Build-input binding), `check
  preconditions` (Vet preflight rows), `check seal` (structure
  plus exact artifact bindings and evidence freshness), `check slice` (slice
  contract preflight before wright dispatch), `check task-graph`, `check
  diff-scope` (mechanical changed-paths ⊆ allowlist before reviewer dispatch),
  `check path-disjoint`, `check skill-trust`, `check indexes`, `check
  regression` (progress high-water mark), `check drift` (readiness-input
  attribution), `check windows` (deferral-marker waivers), `check dup`
  (advisory near-duplicates), `check anchors` (Critical/Important quote
  anchoring), `check perturb` (safe-perturbation probe), `detect commands` (repository test/lint wiring).
- Observation: `observe summary` (`orient` alias), `observe slice`, `next`
  (minimal remaining lifecycle path), `handoff` (deterministic resume record).
- Coordination: `context` emits one deduplicated read-set bundle from each
  skill's `loads:` manifest into `.devrites/work/<slug>/ctx/` — dispatch targets
  read it instead of selecting files. It prints `unselected=[...]` for declared
  triggers not passed; an omitted applicable trigger is a gap, not a shortcut. A
  `--role` call is a dispatch, so the `agents` trigger auto-fires when declared
  (`auto=[agents]` in output); a role with no agent contract file fails instead
  of emitting a contract-less packet.
  Manifest grammar: `always` loads every call; `triggers` maps names to files that
  load only when passed or suggested; `workspace` lists the feature artifacts every
  reader may need; `workspaceByRole` maps a dispatch role to its own artifact list —
  present role wins, absent role falls back to `workspace`, an empty list means the
  role reads no workspace files. The engine prints `suggested=[...]` triggers
  inferred from workspace facts (AFK/parallel flags, frontend/security wording,
  principles presence); confirm them rather than re-deriving conditions. Repeat
  calls with unchanged inputs print `unchanged` and reuse the existing bundle.
  `dispatch` owns the launch-wave barrier — seal needs a distinct handle per
  role (the last start seals), return needs seal, open --claim-session/--claim
  takes the wave's file claim (released by the completing return or abandon
  --release), and start/return auto-record into the `metrics` ledger;
  `candidate freeze` and `packet` write a wave's frozen diff and packets. `claim` owns the advisory session-scoped `claims.jsonl` ledger;
  `note` owns anchored `notes.md` entries (`check seal` refuses non-exact
  anchors); `parallel` owns deterministic worktree lease/create/integrate/
  cleanup for parallel slices.
- `gates` owns the machine-checked acceptance ledger (`gates.md`): `scaffold`,
  `status`, `run`, `reverify`, `lint`, `precheck`, `attest`, `abandon` — grammar and
  authoring rules in
  [`reference/standards/gates.md`](reference/standards/gates.md).
- Atomic state: `state resolve` for answer/drop/batch, `state merge-manifest`
  for the predecessor-chain union, `state set`/`state append` for cursor fields and
  ledger sections, transactional `state close`, and `migrate`
  for fail-closed workspace-schema normalization.
- `secret-scan`, `open-visual`, and `version` remain read-only helpers.
- Offline, local `install`, `update`, and `uninstall`; their shell/npm callers
  acquire the candidate bundle, source, and binary before invoking the engine.

Exact native agents/checklists own semantics; the host filesystem owns ledger
reads/preview/confirmed no-clobber writes, spec grammar re-reading, question-id
allocation, clarify cursor edits, AFK accounting, recovery accounting, and
read-only diagnostics. Repository scripts and CI own gates.

Files in this skill

  • SKILL.md5.7 KB
  • reference/candidate-integrity.md3.4 KB
  • reference/intent-map.md8.7 KB
  • reference/orchestration-profiles.md2.8 KB
  • reference/parallel-dispatch.md12.2 KB
  • reference/reply-contract.md5.4 KB
  • reference/standards/README.md5.7 KB
  • reference/standards/acceptance-preserving-reslice.md3.7 KB
  • reference/standards/afk-hitl.md28.7 KB
  • reference/standards/agents.md16 KB
  • reference/standards/anti-patterns.md4 KB
  • reference/standards/architecture-health.md3.3 KB
  • reference/standards/assumption-checkpoints.md2.1 KB
  • reference/standards/audit-coverage.md5.5 KB
  • reference/standards/browser-proof-checklist.md2.4 KB
  • reference/standards/ci-cd.md3.8 KB
  • reference/standards/code-index.md6.2 KB
  • reference/standards/code-navigation.md3.8 KB
  • reference/standards/code-review.md11.8 KB
  • reference/standards/coding-style.md3 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

View all in development →

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.