Repository navigation
Open
Labels
documentationImprovements or additions to documentationImprovements or additions to documentationepicParent tracking issue coordinating multiple child issuesParent tracking issue coordinating multiple child issues
Description
Summary
These are validated findings from an automated audit (Fable) run against a public clone, confirmed against current main. They focus on two themes:
- First-day default-path experience — what a first-time user on the DEFAULT backend (
msb) actually hits. Several docs steer the reader toward the non-defaultsbxpath, while the real default-path failure modes and constraints are undocumented or buried. - Doc accuracy — factual inaccuracies: undefined terms, dead cross-references, a withdrawn NIST control, a copied-over allowlist that references paths absent here, and prose that no longer matches the code.
Scope note
One related finding — the sbx .git heal-loop functional bug — was already fixed by #327 (the heal probe moved from .git to AGENTS.md). However, the prose sub-claim that the sandbox "clones" the playbook still remains and is tracked as a child issue below.
Child issues
- USAi is never defined and the 'nothing to get in advance' claim contradicts the hard abort #336 — USAi is never defined and the "nothing to get in advance" claim contradicts the hard abort
- Seven references point at a local docs/CODING_PRACTICES.md that does not exist in this repo #337 — Seven references point at a local docs/CODING_PRACTICES.md that does not exist in this repo
- Competing backend-primacy claims, and the default backend's single-agent limitation isn't where the reader chooses #338 — Competing backend-primacy claims, and the default backend's single-agent limitation isn't where the reader chooses
- Troubleshooting/KNOWN_FAILURE_MODES covers the non-default backend; none of the default-path first-day failures are in it #339 — Troubleshooting/KNOWN_FAILURE_MODES covers the non-default backend; none of the default-path first-day failures are in it
- QUICKSTART export block leaks the USAi key into shell history and isn't marked as the non-interactive path #340 — QUICKSTART export block leaks the USAi key into shell history and isn't marked as the non-interactive path
- Quickstart cites withdrawn NIST control SA-12 where the playbook correctly uses SR-3 #341 — Quickstart cites withdrawn NIST control SA-12 where the playbook correctly uses SR-3
- The secret-scanner allowlist was copied from the patterns repo and exempts paths that exist only there #342 — The secret-scanner allowlist was copied from the patterns repo and exempts paths that exist only there
- Prose still says the sandbox 'clones' the playbook after the heal probe moved off .git #343 — Prose still says the sandbox "clones" the playbook after the heal probe moved off .git
Surfaced via an automated audit (Fable) against a public clone; validated against main 2026-08-19. AI-assisted (OpenCode).
Metadata
Metadata
Assignees
Labels
documentationImprovements or additions to documentationImprovements or additions to documentationepicParent tracking issue coordinating multiple child issuesParent tracking issue coordinating multiple child issues
Activity
wz-gsa commentedon Aug 20, 2026
Grooming (2026-08-20) — epic status (first-day default-path experience + doc accuracy)
In open PR #345:
CODING_PRACTICES.mdrefs → canonical playbook URL (fixed 8 of 9; the 9th inchecklists/pre-deployment.mddeferred — collides with open PR docs(controls): replace withdrawn NIST SA-12 with SR-3 (supply chain) #344)read -rs+ interactive cross-linkAlready fixed elsewhere:
Deferred, still open:
pre-deployment.md:11in after PR docs(controls): replace withdrawn NIST SA-12 with SR-3 (supply chain) #344 mergesRecommend keeping open until #339 (needs a KVM host) and #342 land.
docs: fix first-day default-path accuracy (USAi, backend primacy, dea…