Release Gate: v1.0.0

Completed 2026-08-17. Every row records the command run and its result. Rows that cannot apply are marked N/A with the reason, never left blank.

Toolchain currency

Check	Result	Evidence
`hdk` pin matches crates.io stable	PASS	`hdk 0.7.0`
`hdi` pin matches crates.io stable	PASS	`hdi 0.8.0`
`holochain` pin matches crates.io stable	PASS	`holochain 0.7.0`
`@holochain/client` pin matches npm latest	PASS	`0.21.0`, published 2026-07-31
`@holochain/hc-spin` pin matches npm latest	PASS	`0.700.0`, published 2026-07-31
holonix ref exists and is the 0.7 line	PASS	branch `main-0.7`, tip `ffcc7c63b4b87dde16a69247775639b49c5778b1`, matched by the example’s `flake.lock`
kangaroo-electron ref bundles 0.7	PASS	branch `main-0.7`, `kangaroo.config.ts` declares `holochainVersion: '0.7.0'`
Scaffolder guidance names the stable release	PASS	`holochain_scaffolding_cli 0.700.0`, and the skill documents that holonix bundles `0.700.0-rc.0` instead
Companion libraries checked by pin, not tag	PASS	Wind Tunnel tag `v0.7.1` pins hdk 0.6.3 while `main` pins 0.7.0; hREA `happ-0.4.0-beta` pins hdk 0.6.1. Both stated in the skill

Skill integrity

Check	Result	Evidence
Validator passes on a clean tree	PASS	`sh scripts/validate-skill.sh` exits 0
No orphaned markdown	PASS	validator orphan check
Every routing target resolves	PASS	validator routing check
No 0.6-era removed APIs in code blocks	PASS	validator forbidden-API check
Old JS test harness retired to a single pointer	PASS	validator retirement check
No leftover placeholder markers	PASS	validator marker check
Docs site builds	PASS	`mdbook build` exits 0, writes `book/`

Code correctness

Check	Result	Evidence
Example hApp zomes compile to wasm	PASS	`RUSTFLAGS='--cfg getrandom_backend="custom"' cargo build --release --target wasm32-unknown-unknown`, exit 0
DNA packs from freshly built wasms	PASS	`hc dna pack`, exit 0, wrote `forum.dna`
hApp packs	PASS	`hc app pack`, exit 0, wrote `my_forum.happ`
Sweettest suite passes	PASS	`cargo test --release -- --test-threads 4`, exit 0. `create_post`, `create_and_read_post`, `create_and_delete_post`, `create_and_update_post` (4 passed, 35.38s); two-agent `link_a_agent_to_a_post` (1 passed, 10.97s)
Generated-code idiom matches the stable scaffolder	PASS	Example regenerated by `hc-scaffold 0.700.0`; output identical to the committed tree except the two files carrying the rc idiom, both adopted verbatim
Templates match the example	PASS	`assets/templates/integrity-lib.rs` derived from the adopted source, zero `WrongActionError`

Skill efficacy

Check	Result	Evidence
Live end-to-end, run A (2026-08-17)	PASS, with a caveat	GLM 5.3 via opencode, task: add a `Comment` entry type with CRUD, validation and a `PostToComments` link. Wrote 6 files, then `cargo build --release --target wasm32-unknown-unknown` exit 0 on first try. Caveat: described at the time as isolated on the strength of `--pure`. It was not. `--pure` disables plugins only; the `instructions` array still loaded eleven local files. The compile evidence stands; the routing and idiom evidence is weakened, because a context-loaded agent is not the naive reader the check describes
Live end-to-end, run B (2026-08-19)	PASS, with the same caveat	A later, separate run against the iteration-3 content: membrane gating plus a source-chain query, `E2E_WASM_BUILD_EXIT=0`, `posts_integrity.wasm` 1732446 bytes and `posts.wasm` 3123616 bytes. Same isolation defect as run A. These are two distinct runs on two different task sets, not two descriptions of one run
Live end-to-end under verified isolation	OPEN	The isolation recipe (`HOME` + `XDG_CONFIG_HOME` + `XDG_DATA_HOME` overridden onto a sandbox home) was established on 2026-08-19 and verified by behaviour rather than self-report. Re-running runs A and B under it is tracked as C46 and C47 in the project ISA. Not a 1.0 blocker: the compile evidence is unaffected, and what is open is how much the skill alone contributed to the routing
End-to-end output uses current idiom	PASS	`try_from_action` present, zero `WrongActionError`, `LinkQuery::try_new` and `GetStrategy` used, zero removed-in-0.7 APIs
Routing regression guard	PASS	`scripts/eval/run-eval.sh`, 27/39 (69.2%) against a 65% floor. This is a lexical coverage guard, not an accuracy metric. See `scripts/eval/README.md`
Multi-conductor network test on real infrastructure	N/A	Requires a deployed bootstrap and relay. Out of scope for a documentation skill; the two-agent Sweettest covers DHT propagation in-process
Cross-tool portability (other AI clients)	N/A	Deferred to v2 in `docs/requirements.md`. v1.0 targets Claude Code. The cross-model checks that were run are the two GLM 5.3 end-to-ends above
Windows support	N/A	Holochain scaffolding refuses to set up Nix on Windows (`ScaffoldError::NixSetupError`). Upstream limitation, not a skill gap

Community readiness

Check	Result	Evidence
License present	PASS	`LICENSE`, Apache-2.0
Contribution guide present and accurate to this repo	PASS	`CONTRIBUTING.md`, written against the actual scripts and layout
Code of conduct present	PASS	`CODE_OF_CONDUCT.md`, Contributor Covenant 2.1
Issue templates present	PASS	bug report, feature request, stale version report, plus `config.yml`
PR template present	PASS	`.github/pull_request_template.md`
CI runs the validator	PASS	`.github/workflows/validate.yml`
Requirements open questions resolved or explicitly deferred	PASS	`docs/requirements.md` now has a Decisions section; the two that remain are flagged as maintainer calls that do not gate v1.0
No personal or PAI-specific content in the public tree	PASS	`MEMORY/` and `references/example-happ/target/` are gitignored. No absolute home-directory paths, personal identity, or private-tooling references in tracked files. The `~/.claude/skills/` paths in `README.md` are install instructions for a Claude Code skill, which is what the repo is