public/, core/, app/, db/, scripts/, tests/, applicationhost.config, run_site.cmd, TESTING.md, logs/ (removed); .devfoundry/references/environments.md (reverted to placeholders); README.md, CLAUDE.md (added template-hygiene note)DRAFT before user approval.specs/active/ was empty; no conflict.db/ was removed entirely, not altered — see Technical Design.completed: timestamp space.SPEC-004 vendored a full RouteKit application (with a real, working .accdb, IIS Express config, and passing test suite) directly into the DevFoundry repo root to prove the spec-driven process could actually build and verify real application code, not just documentation. Having proven that, the user pointed out that DevFoundry is meant to stay a reusable process template — vendoring one specific app instance into it defeats that purpose. This spec removed the vendored application and restored DevFoundry to an app-free scaffold, while keeping the documentation of RouteKit's conventions (.devfoundry/references/asp-classic-framework.md) and the standing decision that RouteKit is the sanctioned ASP Classic framework for projects that adopt DevFoundry — those are reusable across projects; the vendored code itself was not.
No other active specs existed, so there were no target-file conflicts.
public/, core/, app/, db/, scripts/, tests/, applicationhost.config, run_site.cmd, TESTING.md, and logs/ no longer exist in the repo. — PASS.devfoundry/references/environments.md contained real values specific to the now-removed app instance, when this spec is applied, then those sections are reverted to blank placeholders. — PASS.devfoundry/references/asp-classic-framework.md documents RouteKit's conventions in the abstract, when this spec is applied, then that file is preserved unchanged. — PASSREADME.md or CLAUDE.md, then a short note explains DevFoundry should stay application-code-free. — PASSdb/webdata.accdb (containing only schema_migrations and users tables, per SPEC-004's own schema check) was deleted along with the rest of the vendored tree. No data of consequence existed in it. No rollback beyond re-running SPEC-004's vendoring steps is needed if ever wanted again.find . -maxdepth 1 — RouteKit files present (public, core, app, db, scripts, tests, applicationhost.config, run_site.cmd, TESTING.md, logs), environments.md containing real values — files: this spec — completed: 2026-08-19 23:06public/, core/, app/, db/, scripts/, tests/, applicationhost.config, run_site.cmd, TESTING.md, logs/ — files: as listed — completed: 2026-08-19 23:07.devfoundry/references/environments.md Runtime Versions, Build/Test/Lint Commands, and Legacy Error Log real-value fields back to blank placeholders (kept a pointer to asp-classic-framework.md's reusable IIS Express/aspunit findings so a future project doesn't have to rediscover them) — files: .devfoundry/references/environments.md — completed: 2026-08-19 23:08README.md and CLAUDE.md — files: README.md, CLAUDE.md — completed: 2026-08-19 23:08environments.md has no residual app-specific real values, asp-classic-framework.md unchanged — files: repo root, .devfoundry/references/ — completed: 2026-08-19 23:09find . -maxdepth 1 after removal shows only .abacusai, .devfoundry, AGENTS.md, CLAUDE.md, README.md, docs, specs — none of the removed paths remain.grep for the previous real Data Source/ErrorLogPath values in environments.md returns nothing; Runtime Versions and Build/Test/Lint Commands are blank again.asp-classic-framework.md was not touched in this spec's implementation plan — confirmed no edit was made to it (only environments.md, README.md, CLAUDE.md were edited, plus the vendored-tree removal).grep -n "application-code-free\|process template, not a project repo" README.md CLAUDE.md returns a match in both files.specs/archive/SPEC-004-routekit-bootstrap.md was not modified — archived specs remain immutable history describing a past, since-reverted state.Filled by the Process Improver per .devfoundry/skills/process-retrospective.md, after QA approval and before archiving.
.devfoundry/skills/spec-interview.md step 2 (“identify which stack(s) are affected”) and the Solution Architect persona don't currently prompt the Requirements Analyst to ask “should this actually be vendored into the DevFoundry repo itself, or does it belong in a separate project repo that adopts DevFoundry?" before a spec like SPEC-004 starts touching real framework code. That question would have caught this one spec earlier before real work was reverted. Not yet escalated (first occurrence of this specific gap) — noted here so a second occurrence triggers the 3-count rule via the Friction Tally..devfoundry/improvement-log.md: yesPowered by TurnKey Linux.