All plugins

locksift

Exact, local review of dependency lockfiles: npm package-lock.json, pnpm-lock.yaml, yarn.lock (v1 and Berry), uv.lock, poetry.lock, Pipfile.lock, Cargo.lock, composer.lock, Gemfile.lock and go.sum. lock_diff compares two versions of a lockfile (a branch against main) and returns the dependency changes instead of thousands of lines of churn, with supply-chain signals per installed copy: a source that moved (registry to git or URL, another registry host, a different tarball or commit at the same version, an alias or patch that changed, a tarball that belongs to another package), checksums changed, removed or weakened, new install scripts, new registries. lock_summary gives counts, packages at several versions, non-registry sources and install scripts; lock_why shows the dependency chains that pull a package in. Standard library only: no network, no credentials, no subprocess (never runs git or a package manager), no third-party dependencies, no self-update. Disclosure: reads the local lockfile paths you pass it (read-only), only regular files, and honours Hermes' core read denylist.

rodricksz4h5Toolsv1.0.0Updated 5 days ago

About

Exact, local review of dependency lockfiles: npm package-lock.json, pnpm-lock.yaml, yarn.lock (v1 and Berry), uv.lock, poetry.lock, Pipfile.lock, Cargo.lock, composer.lock, Gemfile.lock and go.sum. lock_diff compares two versions of a lockfile (a branch against main) and returns the dependency changes instead of thousands of lines of churn, with supply-chain signals per installed copy: a source that moved (registry to git or URL, another registry host, a different tarball or commit at the same version, an alias or patch that changed, a tarball that belongs to another package), checksums changed, removed or weakened, new install scripts, new registries. lock_summary gives counts, packages at several versions, non-registry sources and install scripts; lock_why shows the dependency chains that pull a package in. Standard library only: no network, no credentials, no subprocess (never runs git or a package manager), no third-party dependencies, no self-update. Disclosure: reads the local lockfile paths you pass it (read-only), only regular files, and honours Hermes' core read denylist.

locksift

Exact, local review of dependency lockfiles, as three Hermes tools.

A one-line dependency bump can rewrite thousands of lockfile lines: entries
reorder, hashes churn, packages move between nesting levels. That diff is
unreadable in review, and it is exactly where a supply-chain change hides: a
package that now comes from a git URL, a checksum that changed under the same
version, a new install script. locksift reads both versions of the lockfile and
returns the dependency changes themselves.

ecosystemlockfiles
npmpackage-lock.json / npm-shrinkwrap.json (lockfileVersion 1-3, both sections of v2), pnpm-lock.yaml (5.x, 6, 9, including pnpm 11+'s env document), yarn.lock (v1 and Berry)
Pythonuv.lock, poetry.lock (1.x and 2.x), Pipfile.lock
RustCargo.lock
PHPcomposer.lock
RubyGemfile.lock
Gogo.sum

The format is detected from the content. locksift never runs git or a package
manager: you pass the files. To get the base version of a lockfile:

git show origin/main:package-lock.json > /tmp/base-package-lock.json

Why a tool and not a prompt or skill

  • The real change, not the text change. Real output on two versions of
    Hermes' own uv.lock (354 packages, 9.6 MB):

    {"counts": {"added": 0, "removed": 1, "upgraded": 5, "major_upgrades": 2},
     "upgraded": [{"name": "filelock", "from": "3.32.7", "to": "4.2.6", "major": true},
                  {"name": "setuptools", "from": "83.0.0", "to": "84.1.0", "major": true},
                  {"name": "tornado", "from": "6.5.10", "to": "6.6"}, ...],
     "removed": [{"name": "tenacity", "versions": ["9.2.1"]}],
     "signals": {}}
    
  • Supply-chain signals. Each installed copy is compared with its
    counterpart (the same install path, or the same name and version), never with
    a pool of all copies. signals lists:

    • source_changed: what changed (kind, host, alias or ref) and from
      what to what. That covers registry to git or a URL; another registry host,
      including a package moving from a private registry to the default one at a
      new version; a different tarball, git commit or URL at the same version; and
      an alias ("foo": "npm:bar@1") that now names a different real package, in
      npm, pnpm and Yarn.
    • same_version_checksum: the same version with a changed checksum, a new
      artifact (a uv wheel, a Poetry file or a Pipfile hash), a removed checksum,
      or a weaker algorithm (sha512 to sha1).
    • new_install_scripts: a package that runs install scripts and did not in
      any earlier version. This includes a new Composer plugin. Formats that do
      not record scripts are named in notes.
    • new_non_registry: new git, URL and local-path packages.
    • new_registry_hosts: registries the lockfile never used before, including
      sources a Pipfile or Gemfile declares.
    • sections_disagree: npm lockfileVersion 2 keeps two copies of the tree:
      packages for npm 7+ and dependencies for npm 6. Their versions, URLs
      or checksums disagree.

    source_changed also covers:

    • a local patch added or changed (pnpm patchedDependencies, Yarn patch:);
    • a registry package replaced by a local project of the same name (a Cargo
      [patch] path, an npm link, a uv editable);
    • a git dependency whose commit changed. The same commit spelled differently
      (github:o/r#sha, git+ssh://…/o/r.git#sha, a codeload URL) is not a change.

    It does not cry wolf on churn:

    • A format change is not a checksum change: sha1 becoming sha512, or a new
      Yarn cache key.
    • A field only one lockfile records is not compared, so moving from Yarn v1
      to Berry is quiet.
    • registry.yarnpkg.com and registry.npmjs.org are the same registry, as
      are crates.io's two index URLs.
    • On 74 real version-to-version diffs of the lockfiles in the test corpus, the
      only signals are two real ones: a pnpm patch and a Yarn patch added by the
      project.

What it does not detect

A quiet diff means none of the signals above fired. It does not mean nothing
risky changed:

  • Compromised code published to the registry under a new version looks like
    an ordinary upgrade. Checksums vouch for bytes, not intent.

  • Registries are compared by host. A different project or tenant on the same
    host (two GitLab package registries, for example) is not told apart.

  • Some formats do not record install scripts: pnpm 9, Yarn Berry, uv, Poetry,
    Pipfile, Cargo build scripts and go.sum. notes names them.

  • Cargo.lock does not distinguish a workspace member from a path dependency, so
    a brand-new path crate is not reported (a registry crate replaced by one is).

  • Only the lockfile is read. A change made only in package.json,
    pyproject.toml or .npmrc (a registry setting, say) is outside it.

  • Why is this installed? lock_why follows the lockfile's own dependency
    graph, with npm's nested node_modules resolution, pnpm peer suffixes,
    Yarn ranges and PEP 503 names. It returns the chains from a direct
    dependency or workspace project down to a package:

    {"version": "3.20", "required_by": [{"by": "httpx@0.28.1"}, {"by": "requests@2.34.2"}, ...],
     "chains": [["project hermes-agent", "httpx@0.28.1", "idna@3.20"], ...]}
    

Tools

All three always return a JSON string. A bad path or a file that is not a
lockfile returns { "ok": false, "error": "..." } rather than raising.

lock_summary

argtyperequireddefaultmeaning
pathstringyes-lockfile
max_itemsintegerno50entries per list (cap 1,000)

Returns:

  • packages (unique name and version) and names.
  • direct_dependencies and workspace_projects, when the lockfile records them.
  • dev_only.
  • multiple_versions: names locked at several versions.
  • non_registry: git, URL and local-path packages, with their host and ref.
  • registry_hosts: installed copies per host.
  • aliases, install_scripts, and registry_without_checksum.
  • For go.sum, go_mod_only: modules recorded only for their go.mod.

lock_diff

argtyperequireddefaultmeaning
basestringyes-the older lockfile
headstringyes-the newer lockfile (same ecosystem)
max_itemsintegerno50entries per list (cap 1,000)

Returns:

  • counts.
  • added, removed, upgraded and downgraded, each with a major flag when
    the breaking part of the version changed (X for X.y.z, 0.Y for 0.y.z, 0.0.Z
    for 0.0.z).
  • versions_changed: names locked at several versions where the set changed.
  • direct: added and removed direct dependencies, when recorded.
  • signals.

Version order follows:

  • semver for npm, Cargo, Composer and Go: 2.0.0-1 < 2.0.0, and
    1.0.0-alpha < 1.0.0-alpha.1 < 1.0.0-beta. A Go pseudo-version is a
    pre-release of its base version.
  • PEP 440 for Python: 1.0rc1 < 1.0 < 1.0.post1 < 1.0.1.
  • RubyGems for Ruby.

A version that is not a version, such as a git commit, has no direction: the
change is listed under versions_changed.

lock_why

argtyperequireddefaultmeaning
pathstringyes-lockfile
packagestringyes-package name (@babel/core, requests, serde, ...)
versionstringno-only this locked version
max_pathsintegerno5chains per version (cap 50)

Returns every locked version of the package, each with:

  • source and required_by (the dependent and the requested range);
  • the shortest chains from project <name> (a direct dependency) down to it;
  • direct_in, when it is a direct dependency.

Some lockfiles do not record the project's direct dependencies: npm v1, Yarn,
Poetry and Composer. There a chain starts at a package nothing else requires,
and notes says so. Pipfile.lock and go.sum have no dependency graph, so
only whether the package is locked is returned.

Safety

  • Read-only. It reads only the lockfile paths you pass, and only regular files
    up to 256 MiB.
  • No network, no credentials, no subprocess (it never runs git or a package
    manager), no background work, no self-update. Standard library only
    (json, tomllib). pnpm and Berry YAML is read by a small reader that
    accepts only the YAML those tools write: anchors, tags and multi-line
    scalars are refused.
  • Checksums are compared, never echoed in full.
  • Honours Hermes' core read denylist (credential stores, project .env files,
    NT/device namespace paths). The path is checked as given and as an absolute
    path before anything is opened. Device, fd and /proc paths are refused,
    also through symlinks.

Install

hermes plugins install rodricksz4h5/locksift --enable

License

MIT

README from the reviewed commit 1f09443; it updates when the author re-pins.

More by rodricksz4h5

8 plugins by rodricksz4h5 in the catalog.

rodricksz4h5covsiftExact, local analysis of code coverage reports: LCOV (c8, nyc, Jest, Vitest, cargo-llvm-cov, lcov), Cobertura XML (coverage.py, Jest/Vitest, coverlet, gcovr, PHPUnit), JaCoCo XML, Clover XML, SonarQube generic XML, coverage.py, gcovr and Coveralls JSON, and Go cover profiles, from one file or a directory of them. coverage_diff takes a unified diff (git diff) and returns the added lines no test ran, per file, with diff coverage and an optional fail_under gate; coverage_summary gives exact line, branch and Go statement coverage, the least covered files with their uncovered lines and a per-directory rollup; coverage_file lists one file's uncovered lines and partial branches; coverage_compare shows which files lost coverage between two runs. Standard library only: no network, no credentials, no subprocess (never runs git or tests; the diff is passed in), no third-party dependencies, no self-update; XML entity declarations are refused. Disclosure: reads the local report and diff paths you pass it (read-only), only regular files, and honours Hermes' core read denylist.Communitytoolsrodricksz4h5dataprobeExact, local structure and quality checks over large CSV/TSV and JSONL data files that don't fit in a model's context. data_profile reports per-column inferred type, null rate, distinct count, numeric statistics and top values (for JSONL, the inferred union schema with per-field presence); data_validate reports only the actionable problems — mixed-type columns with the offending line numbers, ragged rows, duplicate rows, non-unique key candidates, constant or all-empty columns and malformed JSONL lines; data_head returns a clean parsed preview that handles quoted CSV and JSON correctly. Streamed, standard library only: no network, no credentials, no third-party dependencies, no self-update. Disclosure: reads the local data path you pass it (read-only) and honours Hermes' core read denylist.Communitytoolsrodricksz4h5layersiftExact, local reading of saved container images: docker save and podman save tars (classic and Docker 25+ layouts) and OCI image archives, including multi-platform indexes and gzip, bzip2 or xz compression. image_summary says where the size goes: each layer with the Dockerfile instruction that made it, the final filesystem size, bytes still shipped after a later layer deletes or overwrites them (which layer added them, which removed them), caches and archives left behind (apt lists, apk/yum/dnf, pip/uv/npm/yarn, Go and Cargo caches, temp directories, .git, compilers) with the fix, and the biggest directories and files; image_layer opens one layer; image_compare compares two images before and after a Dockerfile change. Standard library only: no network, never runs docker or contacts a registry, no credentials, no subprocess, never extracts to disk, no third-party dependencies, no self-update. Disclosure: reads the local image archives you pass it (read-only), only regular files, and honours Hermes' core read denylist.Communitytoolsrodricksz4h5logmineExact, local intelligence over large text log files. log_summarize mines the file into recurring message templates (masking numbers, IPs, UUIDs, hashes, quoted strings and timestamps so near-identical lines collapse into one ranked pattern with an exact count), plus a level histogram and the detected time span; log_grep is regex search with line numbers and grep-style context; log_timeline buckets events per minute/hour/day to find spikes. Streamed, so a log too big to fit in the model's context still works. Standard library only: no network, no credentials, no background processes, no self-update. Disclosure: reads the local log path you pass it (read-only).Communitytools

Developer Mode

Explore built-in Hermes toolsets and actions.

HermesAgent

Developer?

Create a Hermes extension for Desktop or agent workflows.

Developer Section