Repository navigation
design(atlas): source-mark grammar — one parser + one formatter for composite provenance marks #6044
Description
Parent: #6018 (Atlas ingestion program tracker). Research source: prior-art doc ~/workspace/scratch/atlas-ingestion-brief-2026-11-04/prior-art-research.md (2026-11-04). Follow-up 16 (Cluster 3). Implementation home: the provenance registry #3125; audit consumer: #6029.
Finding (research, in substance)
The brief's marks — repo@hash:path and "OpenCode DB @ AMD laptop on this day" — are composite marks, and proven grammars exist to steal: git blob/commit SHAs (content identity + tree state), Go module pseudo-versions (v0.0.0-20261104120000-abcdef123456 — compact time+commit at ecosystem scale), OCI digests (name@sha256:… — the locator@digest shape). For non-git sources the mark needs the tuple (system, host, locator, captured-at, content-hash) — Drew's prose mark is that tuple; the work is making it one parseable grammar with one formatter, so the existing-embed audit (#6029) is a parse-and-validate job, not a regex safari.
Scope
- Grammar covering:
repo@commit:path,system@host@date(session stores, DBs),chat@session@timestamp, plus the general fallback form. - The content hash rides in the mark (a
repo@commit:pathmark must stay resolvable even if the repo is later lost). - Validation rules the audit can execute mechanically: parseable? fields in range? hash verifies against stored content?
Done looks like
A grammar spec (EBNF or equivalent) + formatter/parser contract + the validation rule list #6029 consumes.
Activity
developing-today commentedon Nov 5, 2026
CODE-1709
dezren39 commentedon Nov 5, 2026
🤖 Agent note · done
Triage findings (verified 2026-11-05 against fresh GitHub state)
repo@commit:path,system@host@date,chat@session@timestamp, plus the general(system, host, locator, captured-at, content-hash)fallback), a parser/formatter contract, and the mechanical validation rule list. The research basis already exists (2026-11-04 prior-art doc named in the issue body: git SHAs, Go pseudo-versions, OCI digests as the proven shapes to steal).