Centralize YAML handling with safe unmarshal wrappers

[jeduden]

Summary

This PR extracts YAML security and marshaling logic into a new internal/yamlutil package, centralizing all user-content YAML parsing under safe unmarshal wrappers that prevent billion-laughs denial-of-service attacks.

Key Changes

• New internal/yamlutil package with three main functions:

  • UnmarshalSafe() — unmarshal user content into Go structs/maps with anchor/alias rejection
  • UnmarshalNodeSafe() — unmarshal into raw yaml.Node trees with the same security checks
  • Marshal() — thin wrapper around yaml.Marshal for consistency
  • RejectYAMLAliases() — core security check that decodes YAML into nodes without expanding aliases, safely detecting billion-laughs payloads

• Migrated RejectYAMLAliases from internal/lint to internal/yamlutil and removed the old implementation

• Updated all user-content unmarshal sites to use yamlutil.UnmarshalSafe instead of direct yaml.Unmarshal:

  • internal/config/load.go — config file parsing
  • internal/corpus/config.go — corpus config parsing
  • internal/lint/frontmatter.go — front matter parsing
  • internal/rules/requiredstructure/rule.go — directive and schema parsing
  • internal/rules/catalog/rule.go — catalog front matter parsing
  • internal/archetype/gensection/parse.go — generated section YAML parsing
  • cmd/mdsmith/main.go — front matter reading

• Updated marshal sites to use yamlutil.Marshal for consistency:

  • internal/kindsout/kindsout.go

• Comprehensive test coverage in internal/yamlutil/yamlutil_test.go with 20+ test cases covering:

  • Clean YAML acceptance
  • Anchor/alias rejection (including billion-laughs chains)
  • Edge cases (quoted strings, comments, block scalars, multi-document YAML)
  • Error handling and empty input

• Updated test in internal/lint/yamlsafe_test.go to verify anchor/alias rejection through the front matter parsing path

• Marked plan item 121 complete in plan/121_yaml-handling-review.md

Implementation Details

The security model relies on decoding YAML into yaml.Node without expanding aliases (which prevents the exponential expansion attack), then checking for anchor/alias nodes in the tree. This approach safely detects malicious YAML before the standard yaml.Unmarshal call that would expand aliases. Non-anchor syntax errors are allowed through to be handled by the caller's unmarshal logic.

https://claude.ai/code/session_01WnJMsRhRBvtMHQX1seKaQW

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.58%. Comparing base (2b749c4) to head (7535ba1).
⚠️ Report is 3 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #225      +/-   ##
==========================================
+ Coverage   94.51%   94.58%   +0.07%     
==========================================
  Files         136      136              
  Lines       15850    15839      -11     
==========================================
+ Hits        14981    14982       +1     
+ Misses        530      524       -6     
+ Partials      339      333       -6     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull request overview

This PR centralizes YAML parsing/marshaling behind a new internal/yamlutil package so mdsmith can apply one consistent alias-rejection policy anywhere it reads user-controlled YAML. It fits into the codebase as a security-hardening and cleanup pass across config loading, front matter handling, directive parsing, and YAML-emitting command output.

Changes:

• Added internal/yamlutil with shared safe-unmarshal, node-unmarshal, marshal, and alias-rejection helpers plus dedicated tests.
• Replaced scattered direct YAML unmarshalling at user-input call sites with yamlutil wrappers across config, front matter, directives, corpus config, and CLI code.
• Removed the old internal/lint alias-check helper and marked the YAML-handling plan item complete.

Reviewed changes

Copilot reviewed 15 out of 15 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
plan/121_yaml-handling-review.md	Marks the YAML-centralization plan complete and updates acceptance criteria.
internal/yamlutil/yamlutil_test.go	Adds focused tests for alias rejection and wrapper behavior.
internal/yamlutil/yamlutil.go	Introduces the new shared YAML safety/marshal helpers.
internal/rules/requiredstructure/rule.go	Switches schema/directive/front-matter parsing to the new helpers.
internal/rules/catalog/rule.go	Uses the centralized safe unmarshal for catalog front matter.
internal/lint/yamlsafe_test.go	Retargets alias-rejection coverage through front-matter parsing.
internal/lint/yamlsafe.go	Removes the old alias-rejection implementation after extraction.
internal/lint/frontmatter.go	Routes front-matter kinds: parsing through yamlutil.
internal/kindsout/kindsout.go	Uses the shared marshal wrapper for YAML output.
internal/corpus/config.go	Uses safe unmarshal for corpus config and local overrides.
internal/config/load.go	Replaces config parsing/key inspection with centralized YAML helpers.
internal/config/convention.go	Uses node-safe parsing for raw convention: validation.
internal/archetype/gensection/parse.go	Uses safe unmarshal for generated-section YAML bodies.
cmd/mdsmith/main.go	Uses centralized YAML helpers for front matter reading and init output.
PLAN.md	Updates the plan catalog entry for plan 121 to completed.

[Copilot]

Pull request overview

Copilot reviewed 15 out of 15 changed files in this pull request and generated 2 comments.

[jeduden]

🔍 Merge Queue — bisecting

A larger batch failed CI. Bisection is isolating the culprit: this run tests up to 2 of 3 candidate PRs on merge-queue/batch-bisect-223-1777815378. View current bisect CI run.

Next: No action needed — you'll be notified when the culprit is isolated or this PR merges.

[jeduden]

🟢 Merge Queue — picked up

This PR is in the queue and will be batched with other queue-labelled PRs.

Next: No action needed — you'll get another comment when CI starts on the batch. View merge queue run.

[jeduden]

🔵 Merge Queue — CI running

Merged into batch branch merge-queue/batch-225-1777815487. View CI run.

Next: No action needed — you'll be notified when CI completes.

[jeduden]

✅ Merge Queue — merged

This PR landed on main via commit 81cc3f5. CI run that validated the merge.

Next: Done — nothing more to do here.