skillmeld 0.4.0__tar.gz

skillmeld-0.4.0/.gitignore

@@ -0,0 +1,19 @@
+.venv/
+__pycache__/
+*.py[cod]
+.pytest_cache/
+.ruff_cache/
+.coverage
+htmlcov/
+dist/
+build/
+*.egg-info/
+.DS_Store
+
+# No AI footprint in the public repo — orientation/planning stay local (in workbook)
+CLAUDE.md
+.claude/
+.impeccable/
+
+# Scratch / reference attachments dropped beside the repo (never shipped)
+/_scratch/

skillmeld-0.4.0/CHANGELOG.md

@@ -0,0 +1,184 @@
+# Changelog
+
+All notable changes to skillmeld are recorded here. The format follows
+[Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and the project uses
+[semantic versioning](https://semver.org/spec/v2.0.0.html).
+
+## [0.4.0] - 2026-09-26
+
+### Added
+
+- skillmeld is on PyPI: `uv tool install skillmeld` (or `pipx install skillmeld`) puts the
+  `skillmeld` command on your PATH. Releases publish from the version tag through PyPI trusted
+  publishing; the sdist and wheel carry only the package, its tests and the top-level docs.
+- The Claude Code plugin installs straight from GitHub: `/plugin marketplace add ifylab/skillmeld`
+  fetches the repository into the plugin cache and the skill runs the engine from there. No clone.
+- `awesome-scout`: build-time discovery over a curated awesome-list (default:
+  VoltAgent/awesome-agent-skills). Every linked repository not already in the catalog, ranked by
+  GitHub stars. Candidates only; membership in the source list stays a hand-made decision.
+- Scans name the executable script files a bundle ships (`core:ships-scripts`, informational).
+  Skills that bundle scripts carried a vulnerability about twice as often in a 2026 study of 31k
+  community skills (arXiv 2601.10338), so the review card can say what will run.
+- Trigger scoring lists `leaky_ids`, the queries that spell out their target skill's compound
+  name and so route trivially, and a `held_out_pass_rate_strict` without them; `eval improve`
+  repeats the warning.
+- `skillsmp-scout` reports the remaining daily quota from the API's headers and stops with a
+  plain message when the quota is exhausted.
+
+### Changed
+
+- Discovery matches a hyphenated name on its parts too, so `weekly-production-review` is
+  reachable from "weekly review" and not only from the exact compound.
+- A host referenced from many files collapses into one finding per host and rule, with a count,
+  instead of one finding per line.
+- SkillSpector findings carry the tool's pattern, explanation and matched text; its JSON has no
+  `message` field, so earlier versions showed only the category. The category fold covers all 18
+  categories of v2.12, an empty scan is reported as a notice, and the adapter is verified against
+  v2.12.0.
+- The claude.ai zip and the `/v1/skills` payload render frontmatter within the Agent Skills spec:
+  `disallowed-tools` and `disable-model-invocation` are left out, named in `warnings`, because an
+  upload refuses a SKILL.md that carries them. The Claude Code and marketplace emits keep them.
+- `emit api` reports no required `anthropic-beta` header: the Skills API, the code execution tool
+  and the Files API are all GA. `beta_headers` is empty and `legacy_beta_headers` names the two
+  identifiers still accepted as opt-ins.
+- The SkillsMP scout pages at the documented maximum of 50 per request.
+- SKILL.md says that REVIEW is the usual verdict for a skill that calls the network or reads
+  files, and names the two Claude Code settings that govern the skill-listing budget.
+
+### Fixed
+
+- A multi-line frontmatter `description` (a `>` or `|` block, a quoted or a plain continuation)
+  is read whole. Earlier versions kept the block marker or nothing, which made such skills
+  unreachable by discovery.
+- A trailing period on a frontmatter `license:` value (`MIT.`) no longer breaks the SPDX id.
+
+### Security
+
+- `anyio` raised to 4.15.1, past two advisories fixed in 4.14.2 (IDNA host-name matching in TLS
+  streams, and process-pool workers blocking on stderr). skillmeld does not use either path; the
+  package arrives through httpx.
+
+## [0.3.0] - 2026-08-19
+
+### Added
+
+- Four core rules close gaps a new taxonomy coverage benchmark surfaced: cron and login/startup
+  persistence, ClickFix-style paste-to-fix lures, and directed false reassurance ("tell the user
+  it is safe"). The benchmark maps every publicly named category of agentskill.sh's threat model
+  to at least one live rule and carries OWASP Agentic Skills Top 10 ids as an advisory
+  cross-reference (the OWASP list is in pre-ratification review, so its ids are never a stored
+  schema key).
+- NVIDIA SkillSpector joins semgrep and gitleaks as a PATH-optional, escalate-only scanner
+  adapter. It always runs `--no-llm` (scanned content never leaves the machine; its supply-chain
+  check may send dependency names — never contents — to OSV.dev, with a bundled fallback), its
+  CRITICAL findings cap at REVIEW like every adapter, and its categories fold onto the gate's
+  taxonomy.
+- The weekly catalog build now consults scancode-toolkit for license texts the lightweight
+  fingerprints cannot identify, so gold-standard SPDX detection is baked into the signed catalog
+  while the client stays dependency-light.
+- `emit marketplace` gained `--marketplace-version` and `--owner-url`. The manifest now matches
+  the shape Claude Code marketplaces ship (a `metadata` wrapper with description + version), and
+  the version also lands on the plugin entry so `claude plugin update` can see a re-composition.
+- The SkillsMP adapter is real: `skillsmp-scout --queries ...` runs authed, budget-capped,
+  paginated breadth discovery over the SkillsMP registry and prints candidate `owner/name`
+  repos ranked by stars. Candidates only — catalog membership stays a hand-curated decision —
+  and the undocumented response schema is pinned by fixture so upstream drift fails loudly.
+
+### Changed
+
+- A scan that runs without an optional scanner now says so with a visible notice instead of a
+  quiet `absent` version entry; a security gate announces reduced coverage.
+- `scan --license --sources` distinguishes the two license-unknown cases: the catalog also has
+  no SPDX for the source (unknown is the settled state) versus the bundle simply not being in
+  the provided sources.
+- The merge plan's support-file reference warnings collapse into one aggregate line listing
+  every referenced file, so boilerplate cannot drown a real warning.
+- `eval --help` groups shared, run-only, and improve-only flags; `catalog --help` documents the
+  trust model (verification at sync/verify time; `discover` trusts the last verified sync).
+- `eval run --write-evals` keeps the caller's query numbering whenever the ids carry unique
+  digits, instead of always renumbering the exported cases.
+- Quality-gate body warnings cite the line numbers of unescaped html-like tags.
+- Spec refresh: the Skills API is GA (`skills-2025-10-02` and `code-execution-2025-08-25` are
+  optional opt-ins now; `files-api-2025-04-14` is still required when files move), and Claude
+  Code renamed its routing-budget setting to `skillListingMaxDescChars` (same 1536 default) —
+  constants, warning text, and docs updated.
+
+- The quality gate no longer hard-fails a skill whose body carries an unescaped html-like tag;
+  the finding surfaces as a warning instead. Hard issues now cover only what the composition
+  itself authors (name, description, frontmatter) — composed bodies are byte-traced from source
+  skills and improves are description-only, so a body finding had no in-engine remediation and
+  blocked `eval run` from ever reaching `passed: true` on an affected set.
+
+### Security
+
+- `cryptography` raised to 50.0 — earlier versions expose a timing oracle in PKCS#7
+  decryption, an API skillmeld never calls.
+
+## [0.2.0] - 2026-07-26
+
+### Added
+
+- The hosted data layer is live: `catalog sync` now works out of the box against
+  `data.ifylab.dev/skillmeld` — an Ed25519-signed manifest verified against an embedded public
+  key, a hash-pinned catalog, and an advisory verdict index in which every published skill is
+  pre-scanned (known-BLOCK bundles are dropped at discovery before anyone sees them). Rebuilt
+  weekly by CI.
+- `build-catalog` builds and signs the production artifacts (crawl, fetch-verify, scan, sign)
+  with the signing key from `SKILLMELD_SIGNING_KEY`, and `catalog sync --base-url` points the
+  client at an alternate hosted endpoint.
+- `emit api` output now carries the pinned Skills API beta headers (`beta_headers`), the
+  provenance text with a sharing-scope section (`provenance_md`), a standing warning that a
+  `/v1/skills` upload is workspace-wide, and `requires_confirmation: true` when the merge plan
+  holds a REVIEW frontmatter verdict. It also warns when a composed skill carries support files,
+  which the `/v1/skills` payload cannot include — those travel separately via the Files API.
+- A `marketplace` emit surface that packages the merged set as a `strict:false` Claude Code plugin
+  marketplace (`.claude-plugin/marketplace.json` plus the skills tree and `PROVENANCE.md`), ready to
+  host and install with `/plugin marketplace add`.
+- `emit marketplace --plugin-name` to set the plugin entry's name. Without it, a multi-skill set now
+  defaults to the composed skills' names joined, instead of the generic `orchestrator` slug.
+- `eval` speaks the skill-creator interchange formats: `eval run --write-evals` exports the query
+  set as a portable `evals.json`, `eval improve --history` keeps a `history.json` improvement
+  ledger, and `--ingest-source-evals` reads a source skill's bundled evals as extra train-side
+  trigger queries (never held out, so the leakage gate and the improve selection stay clean).
+
+### Changed
+
+- The catalog crawl resolves every source repo to its commit SHA at build time, so a published
+  catalog's fetch URLs and pinned hashes stay consistent no matter what the branch does
+  afterwards.
+- `emit claude-code` names the provenance file `PROVENANCE-<set>.md`, so two composed sets
+  emitted into one shared skills directory keep separate provenance instead of silently
+  overwriting each other's.
+- README places skillmeld among the newer composition tools (AgentSkillOS, SkillComposer) and
+  links the Agent Skills spec home ([agentskills.io](https://agentskills.io)).
+
+### Fixed
+
+- `eval improve` without `--baseline-judgments`/`--candidate-judgments` returns the CLI's JSON
+  error contract instead of a Python traceback.
+- Passing `ground`'s full output to `--profile` fails loudly with the fix named, instead of
+  silently validating an empty profile that turned pruning into a no-op.
+- Framework detection matches exact package names, so a dependency like `license-expression`
+  no longer misreads as Express.
+- The catalog crawl reads a license file inside the skill's own folder when the repo has none at
+  the root, so a source licensed per-skill no longer resolves license-unknown and drags a whole
+  composed set to unknown.
+- `eval` now accepts `--sources` (parity with `merge` and `emit`), so a source whose `SKILL.md`
+  omits `name:` is verified under its catalog identity instead of failing the byte-trace check.
+- The independent routing cross-check no longer routes near-miss queries on generic programming
+  vocabulary alone ("write", "python", "code", ...), and a token shared by every child carries no
+  routing weight — `independent_trigger` and `routing_disagreements` stay high-precision.
+
+### Security
+
+- `cryptography` raised to 49.0 — wheels before 48.0.1 bundled a vulnerable OpenSSL.
+
+## [0.1.0] - 2026-06-13
+
+First public release. The full pipeline is implemented and tested: intake, repo
+grounding, discovery over a signed catalog, selection (at most three), a tri-state
+security gate, the eight-step byte-traceable merge engine, evaluation, and packaging
+for Claude Code / claude.ai / the API with provenance.
+
+[0.2.0]: https://github.com/ifylab/skillmeld/compare/v0.1.0...v0.2.0
+[0.1.0]: https://github.com/ifylab/skillmeld/releases/tag/v0.1.0

skillmeld-0.4.0/LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

skillmeld-0.4.0/NOTICE

@@ -0,0 +1,5 @@
+skillmeld
+Copyright 2026 Hossein Zargar
+
+This product includes software developed as part of the .ify project
+(https://ifylab.dev).

skillmeld-0.4.0/PKG-INFO

@@ -0,0 +1,144 @@
+Metadata-Version: 2.5
+Name: skillmeld
+Version: 0.4.0
+Summary: Discover existing community skills and merge the best into one tailored, security-scanned skill set.
+Project-URL: Homepage, https://ifylab.dev/tools/skillmeld
+Project-URL: Repository, https://github.com/ifylab/skillmeld
+Project-URL: Changelog, https://github.com/ifylab/skillmeld/blob/main/CHANGELOG.md
+Project-URL: Issues, https://github.com/ifylab/skillmeld/issues
+Author: Hossein Zargar
+License-Expression: Apache-2.0
+License-File: LICENSE
+License-File: NOTICE
+Keywords: agent-skills,claude,compose,merge,skills
+Classifier: Development Status :: 4 - Beta
+Classifier: Environment :: Console
+Classifier: Intended Audience :: Developers
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Software Development :: Libraries
+Requires-Python: >=3.12
+Requires-Dist: bandit>=1.8
+Requires-Dist: cryptography>=43
+Requires-Dist: httpx>=0.27
+Requires-Dist: license-expression>=30
+Requires-Dist: markdown-it-py>=3
+Requires-Dist: pydantic>=2.7
+Requires-Dist: pyyaml>=6
+Description-Content-Type: text/markdown
+
+# skillmeld
+
+<!-- A demo GIF of a real /skillmeld run is planned as the README hero. -->
+
+[![CI](https://github.com/ifylab/skillmeld/actions/workflows/ci.yml/badge.svg)](https://github.com/ifylab/skillmeld/actions/workflows/ci.yml)
+[![PyPI](https://img.shields.io/pypi/v/skillmeld)](https://pypi.org/project/skillmeld/)
+[![License: Apache-2.0](https://img.shields.io/github/license/ifylab/skillmeld)](https://github.com/ifylab/skillmeld/blob/main/LICENSE)
+![Python 3.12+](https://img.shields.io/badge/python-3.12%2B-blue)
+
+Describe what you want to do, point skillmeld at your repo, and it finds existing community skills for the job, security-scans them, and merges the best two or three into one coherent skill set tailored to your project — instead of writing one from scratch.
+
+It runs on your own Claude in Claude Code, grounds in your repo, and shows you what it pulled, what it found, and why before anything is installed. It builds on the existing skills ecosystem (the [open standard](https://agentskills.io), community marketplaces, and registries) rather than replacing it.
+
+## What makes it different
+
+skillmeld composes; it does not generate. Every line in a merged skill traces byte-for-byte back to a source skill — a deterministic verifier enforces this, so the tool can never invent an instruction. The hard, mechanical work (parsing, security scanning, deduplicating, conflict detection, packaging) runs as deterministic Python that makes zero model calls. Your Claude supplies the judgment; the engine supplies the guarantees.
+
+Composition tools are appearing on other layers too: [AgentSkillOS](https://github.com/ynulihao/AgentSkillOS) retrieves skills from a large pool and chains them into runtime workflows, and [SkillComposer](https://arxiv.org/abs/2606.06079) has a model generate and evolve its own skills. skillmeld's job is different: it merges skills other people wrote — untrusted input — into one reviewed, deduplicated set before anything runs, with a security verdict on every source and a byte trace on every line.
+
+## How it works
+
+A Claude Code skill drives a bundled Python engine through one pipeline:
+
+```
+intake -> ground -> discover -> select (<=3) -> fetch -> security gate -> merge -> eval -> emit
+```
+
+- **intake** normalizes the request and says when it is too thin to act on.
+- **ground** scans your repo into a use-case profile, locally.
+- **discover** syncs a signed catalog of community skills (Ed25519-verified, hash-pinned, cached locally) and your Claude ranks the shortlist.
+- **select** takes at most three candidates, a separate cap from the three-skill output.
+- **fetch** downloads only the chosen bundles and checks every file against the hash the catalog pinned.
+- **security gate** scans every candidate (PASS / REVIEW / BLOCK) before you see it, and again after merge. REVIEW is the normal outcome for a skill that calls the network or reads files: the findings are named for you to decide on, and a BLOCK is never overridable.
+- **merge** parses each skill into byte-exact atoms, deduplicates, resolves conflicts, prunes to your use case, and partitions the result into at most three skills behind a thin routing orchestrator. A verifier proves every output atom traces to a source.
+- **eval** scores the set with no model calls: structural quality, byte-traceability, held-out trigger routing and a leakage check; every description edit is gated on it.
+- **emit** packages the result for Claude Code, a claude.ai zip, the API, or a Claude Code plugin marketplace, with a provenance record (`PROVENANCE.md`, or `PROVENANCE-<set>.md` beside a Claude Code skills tree) of where every part came from.
+
+## Install
+
+Two ways in, one engine. The plugin path needs [uv](https://docs.astral.sh/uv/) on your PATH; the command-line path needs uv or pipx.
+
+As a Claude Code plugin, the `/skillmeld` skill, with no clone:
+
+```
+/plugin marketplace add ifylab/skillmeld
+/plugin install skillmeld@ifylab
+```
+
+Claude Code fetches the repository into its plugin cache, and the skill runs the engine from there.
+
+As a command-line tool, from PyPI:
+
+```sh
+uv tool install skillmeld    # or: pipx install skillmeld
+skillmeld --help
+```
+
+The command line runs the deterministic stages one at a time, which suits scripts and CI. The judgment steps (completing the profile, ranking candidates, adjudicating conflicts, authoring descriptions) belong to the skill, driven by your Claude.
+
+## Quickstart
+
+As a skill, describe the use case inside the project it is for:
+
+```
+/skillmeld I review pull requests for a Python service and want one consistent code-review routine
+```
+
+Or exercise individual stages directly from the CLI. Each line stands alone and prints JSON;
+`ground` prints a partial profile whose `summary` and `tasks` you complete before `discover` reads
+it. From the package root of a clone, `uv run skillmeld` is the same command:
+
+```sh
+skillmeld catalog sync                   # fetch and verify the hosted catalog
+skillmeld ground .                       # scan a repo into a profile
+skillmeld scan path/to/skill --license   # security- and license-scan a bundle
+skillmeld merge --bundles a/ b/ --profile profile.json
+```
+
+The full pipeline — `intake`, `discover`, `select`, `fetch`, the eval loop, and every JSON contract these
+commands exchange — is walked step by step in [skills/skillmeld/SKILL.md](https://github.com/ifylab/skillmeld/blob/main/skills/skillmeld/SKILL.md).
+Offline, `skillmeld dev-catalog` builds the same signed catalog locally from repos you name. Two build-time
+scouts feed the curated source list: `skillmeld skillsmp-scout` searches the SkillsMP registry
+and `skillmeld awesome-scout` reads a curated awesome-list; both print candidates, and membership stays a
+hand-made decision.
+
+## What it isn't
+
+- **Not a generator.** It assembles existing skills; it never authors new instructions. A convention no source skill covers is yours to add, not a gap skillmeld fills.
+- **Not a catalog.** It composes from community marketplaces and registries rather than being one.
+- **Not a model.** The engine makes zero LLM calls; the judgment comes from your own Claude, on your tokens.
+
+## Acknowledgements
+
+skillmeld stands on the open [Agent Skills](https://agentskills.io) ecosystem — the skill format, the community marketplaces, and the registries that publish and share skills. It composes that work; it does not replace it. Security scanning leans on [bandit](https://github.com/PyCQA/bandit), with optional [semgrep](https://semgrep.dev/), [gitleaks](https://github.com/gitleaks/gitleaks) and [NVIDIA SkillSpector](https://github.com/nvidia/skillspector) when present.
+
+## Status
+
+In active development, built in the open one piece at a time. The discovery, security, merge, evaluation, and packaging stages are implemented and tested, and discovery runs against a hosted signed catalog rebuilt weekly by CI — every published skill is crawled at a pinned commit and pre-scanned into an advisory verdict index. The curated AEC corpus is coming next. See the [changelog](https://github.com/ifylab/skillmeld/blob/main/CHANGELOG.md).
+
+## Stack
+
+Python 3.12, managed with uv. Ruff for lint and format, ty for type-checking, pytest for tests.
+
+```sh
+uv run ruff check . && uv run ruff format --check . && uv run ty check && uv run pytest
+```
+
+## Contributing
+
+Issues and pull requests are welcome — see [CONTRIBUTING.md](https://github.com/ifylab/skillmeld/blob/main/CONTRIBUTING.md). Contributions are accepted under the project's Apache 2.0 license (inbound = outbound); no separate contributor agreement is required.
+
+## License
+
+Apache License 2.0 — see [LICENSE](https://github.com/ifylab/skillmeld/blob/main/LICENSE) and [NOTICE](https://github.com/ifylab/skillmeld/blob/main/NOTICE).