Stable identifiers and reference integrity

Opportunity: Add repeatable checks for identifiers, internal references, frontier provenance, and supersession links.

Unknowns: Broken-reference rate, duplicate identifiers, and how reliably agents can traverse provenance.

Origin and evidence: ROADMAP.md Repository Fitness; EV-E016; knowledge-platform/Metadata-Standard.md; knowledge-platform/Repository-Architecture.md.

Dependencies: Canonical identifier and link rules.

Method: Machine-readable inventory plus CI validation; seed known failures to calibrate checks.

Outputs and success: Validator, fixtures, remediation report. Success is zero unexplained duplicates/broken canonical references and detection of all seeded defects.

Recommended agent: Repository tooling engineer. Effort: small. Expected gain: reliable substrate for task provenance and future graph work.

Progress (2026-07-30): Implemented the identifier, explicit-link, and frontier-structure validator; calibrated it against seeded defects; and preserved the first clean inventory in data/validation/research-integrity-inventory.json. See research/frontier/investigations/2026-07-30-rfr-009-validation.md.

Progress (2026-08-15): Extended the inventory to measured metadata coverage by artifact class and diagnostic validation of declared related/supersession targets. The baseline covers 131 authored Markdown artifacts: 48 have front matter and 47 have stable identifiers. All five non-empty declared relationships resolve, but no active canonical related or supersession edges exist. Seven seeded/control tests pass. See research/frontier/investigations/2026-08-15-rfr-009-metadata-provenance-coverage.md.

Confidence: High for implemented deterministic checks; medium-low for repository-wide provenance traversal.

Remaining uncertainty: Metadata adoption remains low and no real supersession edges exist to calibrate reciprocal/semantic checks. Heading fragments, bare path references, and registry references are not yet validated. Remain in Validation until a bounded adoption policy and real provenance-edge sample are implemented.