Claude
021dfa6b7b
Real harbor runs reach the Actions log only at job completion (8 quick- subset tasks take 10-30 minutes per agent). Without --debug, the GH log shows zero progress during the run, leaving maintainers guessing if harbor is actually working or stalled. --debug surfaces trial resolution, dataset download progress, and container lifecycle to stdout. Default verbosity already shows trial progress bars; --debug adds the underlying detail. Co-Authored-By: Claude <[email protected]>
heretek
Opinionated Claude Code plugin marketplace: task plugins, quality-gate hooks, curated MCP/LSP/skills.
What is this?
heretek is a Claude Code marketplace that bundles 11 first-party plugins — Rust, Python, JS/TS, web frontend, security, MCP, LSP, skills, agents, output styles — plus the flagship hooks plugin that gates every Edit/Write/MultiEdit with a sub-100ms quality check.
Every item ships after D7 vetting (stars ≥ 500, last commit ≤ 12mo, OSI license, source-audit pass, no critical CVEs in 24 months) with an ADR in catalog/reviews/ documenting the decision.
Install
# Add the marketplace (once)
/plugin marketplace add heretek https://github.com/Heretek-AI/heretek-claude-harness
# Install a plugin
/plugin install rust@heretek
/plugin install hooks@heretek
# Install all 11
for p in rust python js-ts web-frontend hooks security skills-pack mcp-pack lsp-pack agents output-styles; do
/plugin install $p@heretek
done
Common commands
# Inspect local hook event telemetry
python scripts/heretek_cli.py telemetry show
python scripts/heretek_cli.py telemetry grep "Edit"
python scripts/heretek_cli.py telemetry schema
What ships?
See docs/superpowers/specs/2026-08-03-heretek-marketplace-design.md for the canonical plugin catalog. Quick summary:
| Plugin | Category | What it does |
|---|---|---|
rust |
task | rust-analyzer LSP + cargo-check + cargo-clippy skills |
python |
task | ruff LSP + ruff-check skill |
js-ts |
task | biome LSP + typecheck skill |
web-frontend |
task | chrome-devtools-mcp + lighthouse skill |
hooks |
cross (flagship) | 3-layer quality gates (fast blocking <100ms, slow on-demand, git pre-commit) |
security |
cross | security-research + audit-checklist skills (NO hooks — hooks plugin owns all hook components per D15) |
skills-pack |
cross | caveman, superpowers |
mcp-pack |
cross | context7, github-mcp-server, codebase-memory-mcp, serena |
lsp-pack |
cross | rust-analyzer, basedpyright, gopls, clangd |
agents |
cross | code-reviewer, security-reviewer, test-engineer |
output-styles |
cross | terse, detailed, tabular, explanatory |
How vetting works
Every entry in catalog.yaml's items[] carries:
vetting.status(approved | rejected | pending)vetting.date(when last verified)vetting.review(path to ADR incatalog/reviews/)upstream(the actual repo)sha(40-char pin, lock-step with D11 SHA-ride)license(SPDX id)
Quarterly, run python scripts/refresh_pins.py --github-token $GH_TOKEN to re-verify every entry against the D7 bar. Items that fail (stars drop, license drift, critical CVE published) get flagged for review.
Documentation
docs/superpowers/specs/2026-08-03-heretek-marketplace-design.md— design spec (locked decisions D1–D17)docs/recommended-marketplaces.md— external marketplaces to add manuallyCONTRIBUTING.md— how to add new items, submit PRsSECURITY.md— supply-chain risk model + reporting path
Versioning
No version field on first-party plugins (D11). The marketplace itself is versioned by commit SHA. Every plugin entry's sha field pins the exact upstream commit.
License
MIT — see LICENSE.