chore(missing_docs): triage v0.2026.08.19.08.15.stable_01

[rachaelrenk]

Summary

Drift-watch bookkeeping for stable v0.2026.08.19.08.15.stable_01 (first validation run of the missing_docs skill). Every release candidate was triaged against the docs-worthiness criteria; none warranted a feature docs PR. This PR only records verdicts, maps two new factory REST routes as internal, advances the release marker, and refreshes the surface snapshot.

Audit

• check_new_release.py exit 0 — new stable vs v0.2026.08.18.02.52.stable_00
• audit_docs.py --diff exit 0 (audits_skipped empty)
• After map + snapshot update: API findings and surface_changes cleared; only 29 low-severity potentially_stale_docs remain (terminology; owned by style_lint, not drafted here)

Changes

• changelog_decisions.md — six new ledger rows for this release
• feature_surface_map.md — POST /factory/automations/{id}/run and POST /factory/{uid}/runs → internal (Factory REST is x-internal / not in the public OpenAPI publish set)
• last_release_processed.json — marker advanced to v0.2026.08.19.08.15.stable_01
• surface_snapshot.json — regenerated baseline

Triage table (calibration)

Item	Verdict	Gate / disqualifier	Evidence
Warp Factories Early Access (changelog New features)	deferred	Scope: announced major launch	Worthiness criteria exclude major launches from the automated stream; public docs already under src/content/docs/factories/
#14968 inline /usage on Warp Agent CLI	no	Small and intuitive; already documented	Listed on slash-commands.mdx; SlashCommandKind::Usage in static_commands/commands.rs; slash menu surfaces it (Gate 4 cond. 2 fails)
#15181 completions engine memory leak	no	Bug fix restoring intended behavior	No docs described the broken behavior; no knob
#15261 Shift glyph font-fallback (web)	no	Pure UI affordance	Visual-only shortcut hint rendering
#15259 1px workspace border (web)	no	Pure UI affordance	Visual-only
Oz update: MCP server not found after sandbox rebuild	no	Bug fix restoring intended behavior	Restores intended MCP availability; quoted error is the bug being fixed, not a user runbook
POST /factory/automations/{id}/run	no (map internal)	Gate 0 / private API	Not in public OpenAPI; same Factory REST internal pattern as sibling routes
POST /factory/{uid}/runs	no (map internal)	Gate 0 / private API	Same as above
29 stale terminology hits (warp ai, etc.)	deferred	Owned by style_lint	Low-severity; skill says do not duplicate terminology sweeps here

Deferred / not drafted

• No feature PRs this run (successful empty draft set).
• Stale-term findings left for style_lint.
• Snapshot already had changelog_last_version: 2026.08.19 while the release marker lagged, so --diff changelog_review was empty; triage used the live client_version payload + docs changelog instead (skill feedback for humans below in the agent run).

Co-Authored-By: Warp agent@warp.dev

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
docs	Ready	Preview	Aug 24, 2026 7:11pm

[rachaelrenk]

This PR was generated with Warp.

Comment @warp-agent on this PR to send it follow-up work.

 

[warp-agent-staging]

Your GitHub account is not connected to Warp. Connect it here.

[warp-factories]

Your GitHub account is not connected to Warp. Connect it here.

[rachaelrenk]

Closing in favor of a clean re-run after #617.

To be clear about what's being discarded: the triage in here is sound. The six verdicts are well-reasoned — the Gate 4 rejection of #14968 cites SlashCommandKind::Usage in commands.rs and correctly notes the slash menu already surfaces it. It even triaged the three Bug fixes bullets, which the parser deliberately skips, so it covered more than the pipeline requires. The two -> internal Factory run endpoint mappings are correct and independent of changelog triage entirely.

The problem is provenance, not correctness. This run reached those verdicts by manually fetching client_version to work around the marker/snapshot desync that #617 fixes — --diff returned an empty changelog_review, so the pipeline's primary triage input contributed nothing.

Merging it would advance the marker to v0.2026.08.19.08.15.stable_01, which then leaves the post-#617 validation run with nothing to triage. It would no-op and prove nothing about the fix. Since validating #617 end to end is the whole point of the next run, the marker needs to stay where it is.

Everything here regenerates: the verdicts, the surface-map entries, the snapshot. Re-running drift-watch once #617 lands produces the same output through the fixed path, and that run doubles as the validation.

Nothing wrong with the work — just the wrong order.