fix: repair plugin release gate for 0.8.0 (integration tests + evals)#73
Merged
Conversation
1. opencode-plugin 'writes failed AKM tool invocations to OpenCode app logs': stale assertion expected command:"akm"; the error log now records the resolved command (command.displayCommand, the bundled cli.js path) since facd259. Match the resolved command robustly. 2. OpenCode eval harness 'keeps auto-feedback routed through the sandbox akm shim': the 0.8.0 bundled-CLI resolution prefers the real bundled akm-cli over the PATH-installed fake shim, so akm invocations bypassed the shim. Add a default-off AKM_OPENCODE_IGNORE_BUNDLED_CLI opt-out (escape hatch for a broken bundled dep) and set it in the eval sandbox so akm resolves to the deterministic fake shim on PATH. Integration suite: 324 pass / 0 fail. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The plugin invokes curate/search with '--shape agent --format text'. The shim's verb-detection skipped --format/--detail/-q but not --shape, so it parsed the verb as '--shape' and emitted nothing — collapsing curation coverage to 0 in the tier-2 evals. Add --shape to the stripped value-flags; curation recovers to its prior baseline (coverage 0.9125, MRR 0.92). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
akm 0.8.0 made chat.message curation fire-and-forget (no longer blocks the model on the ~8s curate; result injected by system.transform once the async curate resolves). The harness read context synchronously right after chat.message, so context_budget's opencode samples saw 0 chars (drop_rate 1). Poll system.transform until the background curate populates context (bounded 4s). opencode context recovers (avg ~293 chars, drop_rate 0). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Captures the intended 0.8.0 surface (akm_vault tool + akm-vault command removed; env/secret/improve/proposal/memory tools + commands added) and current healthy metrics: curation coverage 0.9125, opencode context restored, feedback precision/recall 1.0, memory secret-leakages 0. Latency captured locally (informational; not gated without --include-latency). Co-Authored-By: Claude Opus 4.8 (1M context) <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.
Repairs the akm-plugin release
testgate, which was red on main (since 2026-06-02) and blocking the 0.8.0 release.Integration tests (2 failures)
command:"akm"; the error log records the resolved command path sincefacd259. Match resolved command robustly.akm-cliover the PATH fake shim, so akm calls bypassed the shim. Added default-offAKM_OPENCODE_IGNORE_BUNDLED_CLIopt-out; eval sandbox sets it.Tier-2 evals (pre-existing breakage)
--shapeglobal flag → mis-parsed verb → curation collapsed to 0. Fixed; curation back to 0.9125 coverage.chat.messagecuration fire-and-forget → context_budget opencode = 0. Harness now pollssystem.transformuntil the async curate resolves.Local gate: integration 324 pass / 0 fail; tier-2 evals exit 0.
🤖 Generated with Claude Code