Host-aware setup, eval instrument repair, and dated support summary (4.2.0) - #104
Merged
Conversation
…rt summary Implements the September 6 audit against 4.1.1: - F1: the activation helper moves into the skill package as scripts/activation.py, resolves the effective trusted file per host (AGENTS.override.md precedence, CODEX_HOME, CLAUDE_CONFIG_DIR, rules/), moves stray blocks, reads plugin inventories for availability, and names managed policy files. A setup playbook lets an installed user ask the agent to enable, check, or disable default governance with one confirmation and configured/available/loaded reported separately. - F3: the canonical large-programme case names catalog-integration-ready, evals/preflight.json registers expected pre-session state and capture_eval.py prepare refuses mismatches, the continuity oracle bans outcomes rather than implementations, and scripts/grade_catalog.py grades retained end states independently of any implementation. - F4: integration.md chooses failed-merge recovery by consequence. - F5: one dated support summary in docs/evidence.md; README, guide, FAQ, SECURITY.md and the site link to it and use effects-based authority text. Version 4.2.0. Host receipts for the exact package are reset to UNVERIFIED until they land. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…ment - An empty AGENTS.override.md is skipped, as Codex skips empty files, and left untouched; AGENTS.md stays the effective file. - Claude unconditional rules count as read by the host, so a block there is configured; rules with paths: frontmatter are shadowed; linked files are inspected but never written through. - Plugin enablement is reported as unknown when config.toml or settings.json cannot be parsed, including on interpreters without tomllib. - The catalog grader accepts nested modules inside the catalog package. - prepare validates and snapshots the fixture tree before the preflight probe imports anything, and rejects a probe that changed files. - Preflight tests drive decisions through controlled Git responses instead of creating repositories. - Evidence no longer claims 4.1.0 Claude streams show policy loading or an unreceipted 4.2.0 install count. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Implements the September 6 independent audit against 4.1.1 and prepares release 4.2.0.
What changes
plugins/skiphow/skills/skiphow/scripts/activation.py. It resolves the trusted user instruction file each host actually reads (non-emptyAGENTS.override.mdoverAGENTS.mdinCODEX_HOME;CLAUDE.mdand unconditionalrules/*.mdunderCLAUDE_CONFIG_DIR), moves shadowed blocks, leaves linked files unedited, reports plugin availability from host inventories, and names managed policy files. A newsetupplaybook lets an installed user ask the agent to enable, check, or disable default governance with one confirmation and configured/available/loaded reported separately.catalog-integration-ready;evals/preflight.jsonregisters expected pre-session state andcapture_eval.py preparerefuses mismatches after validating the tree; the continuity oracle bans outcomes rather than implementations;scripts/grade_catalog.pygrades retained end states independently of any implementation.integration.mdchooses failed-merge recovery by consequence.docs/evidence.md; README, guide, FAQ, SECURITY.md, and the site link to it and use effects-based authority text.Evidence
python scripts/check.pypasses locally (deterministic gate, 330 tests).scripts/check_hosts.py: Claude schema and clean install pass on the candidate; the Codex validator is unavailable locally and the Codex install is refused by managed source policy until the package is onmain.UNVERIFIEDinevals/host-smoke.jsonand will be recorded before tagging.Version 4.2.0 (MINOR: new opt-in playbook and helper; recovery rule widened within existing authority).
🤖 Generated with Claude Code