Riekelt Principal Engineer - Setup Guide¶

Source: riekelt/principal-engineer (3⭐, v1.2.4, pushed Sep 2026) Skill: riekelt/principal-engineer (11 installable skills, 120,279 combined installs) Publisher: riekelt (also ships the riekelt/technical-writer plugin) Category: Engineering Discipline / Code Quality / Agent Safety Quality Tier: 🟡 Beta (functional, author-tested, ships evals; no third-party audits - verified Sep 29, 2026)

"Engineering discipline as skills" - the judgment layer for coding agents. One core skill (principal-engineering) holds the hard rules, risk tiers, and routing; ten discipline skills build on it. Distilled conventions: evidence over theory, no silent error swallows, one home per fact, verification as the definition of done, and fixes that never outgrow their trigger. Ships as a multi-platform plugin (Claude Code, Codex, Cursor manifests) with a change-verifier subagent, /verify and /preflight commands, an evals suite, and semantic-release versioning.

Installation¶

# Claude Code plugin
/plugin marketplace add riekelt/principal-engineer
/plugin install principal-engineer@principal-engineer

# Any Agent Skills host (including Hermes Agent)
npx skills add riekelt/principal-engineer

# Manual: point the platform plugin loader at plugins/principal-engineer/,
# or symlink plugins/principal-engineer/skills/* into the agent's skills directory

Roster - All 11 Skills¶

Skill	Installs	Does
principal-engineering	10,973	Core: pre-change checkpoint, hard rules, risk tiers, rule lifecycle
writing-unit-tests	10,972	Unit tests: behavior not implementation, names as claims, determinism, mock policy
verifying-before-done	10,948	Drive the change at its surface, run the proof, distrust green suites, own failing gates
grounding-before-coding	10,942	Map real code and data first - never guess conventions in unfamiliar code
guarding-architecture	10,937	Named, enforced invariants (statement, rationale, guard); violations mean redesign
keeping-one-source-of-truth	10,933	Derive rather than store; absorb duplicates of data, config, and state
testing-changes	10,932	Decide what tests a change owes; bug-regression pattern, discriminating assertions
scoping-changes	10,931	Size fixes to their trigger; decompose instead of descoping
operating-safely	10,928	Destructive-op guards, secrets hygiene, operator-owned process lifecycles
handling-failures	10,925	No-silent-swallows contract for error paths, catch blocks, fallbacks
adding-dependencies	10,858	Exhaust-what-you-have ladder, posture declarations, pin-and-prove updates

Bundled: change-verifier subagent + /verify and /preflight commands.

CorpusIQ Use Cases¶

Use Case	How
Hermes agent quality gates	`verifying-before-done` and `operating-safely` mirror CorpusIQ's verification-discipline and pre-flight gate patterns for cron/automation work
Skill authoring	`keeping-one-source-of-truth` and `guarding-architecture` align with CorpusIQ's skill-audit and consolidation doctrine
Client dev work	`testing-changes` + `writing-unit-tests` for shipping client integrations with proof, not vibes
Docs repo health	`scoping-changes` and `handling-failures` for the daily sweep/cron scripts that must fail loudly

Integration with CorpusIQ¶

CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.

This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.

See what CorpusIQ connects

Limitations / Verification¶

Designed for coding agents first; the discipline rules transfer to any agent workflow via the Agent Skills-format SKILL.md files
No standalone LICENSE file in the repository - the plugin manifest (plugin.json) declares MIT, but the repo root lacks a LICENSE.md (verified Sep 29, 2026). Confirm licensing before commercial reuse
Single-author project (3⭐); install counts are high (~10.9K per skill) but community audits are not published
Verify: npx skills add riekelt/principal-engineer --list shows 11 skills

Security¶

No skills.sh security audits published (verified Sep 29, 2026):

Audit	Verdict
Gen Agent Trust Hub	Not published
Socket	Not published
Snyk	Not published

← Skills Catalog | Skills Marketplace → Powered by CorpusIQ