.agent mirror issue layer is stale: workflows call issue_claim/issue_release/issue_any_claimable which do not exist

[laird]

CLAUDE.md declares mirror parity CRITICAL, but the entire issue layer under .agent/scripts/ has drifted from plugins/autocoder/scripts/. The Antigravity /fix workflow is broken end to end as a result.

The mirror calls verbs its own library does not define

.agent/workflows/fix.md calls three verbs that .agent/scripts/issue-fns.sh never defines:

verb	called at	defined in .agent?
issue_any_claimable	fix.md:320	❌
issue_claim	fix.md:532, fix.md:648	❌
issue_release	fix.md:575, fix.md:681, fix.md:1479	❌

Verified live:

$ source .agent/scripts/issue-fns.sh
  defined:   issue_list
  defined:   issue_get
  MISSING:   issue_claim
  MISSING:   issue_release
  MISSING:   issue_any_claimable

Consequence: Antigravity /fix exits at preflight, every run

fix.md:319-323:

issue_any_claimable
case $? in
  0) ;;  # work exists; fall through
  1) echo "No claimable issues. Nothing to do."; exit 0 ;;
  *) echo "Backend error while checking for claimable issues"; exit 1 ;;

Undefined function → exit 127 → falls to *) → exit 1. The workflow never reaches any issue. This is not a degraded path; /fix on Antigravity cannot run at all.

Second break: --state blocked is unsupported

.agent's _ifns_gh_list passes --state straight through to gh:

$ source .agent/scripts/issue-fns.sh; issue_list --state blocked --limit 2
invalid argument "blocked" for "-s, --state" flag: valid values are {open|closed|all}

That breaks .agent/workflows/list-needs-feedback.md:43 and .agent/workflows/approve-proposal.md:85.

Third: no backend-level blocking-label filter

.agent's inline _ifns_gh_list applies no label exclusion at all — it never had the no:label bug from #57 because it has no filter whatsoever. Per the 9-verb contract, list --state open is supposed to mean "claimable". The command-layer jq/python filters happen to compensate today, but the backend contract is not honoured.

Scope of the drift

file	status
issues-gh.sh	absent from .agent/ entirely
issues-file.py	438 differing lines; lacks claim/release/any-claimable and the bucket layout (3 bucket refs vs 30)
issue-fns.sh	180 differing lines; inline gh impl vs dispatcher
issue-config.sh	4 differing lines

.agent/scripts/issues-file.py was last touched 2026-05-19; the plugins copy has moved on since.

Fix

Sync the issue layer forward — these are meant to be identical shared infrastructure, and the plugins copy is the maintained one (it implements the full 9-verb contract the mirror's own workflows already assume):

• copy issues-gh.sh into .agent/scripts/
• replace .agent/scripts/issue-fns.sh with the dispatcher version
• refresh .agent/scripts/issues-file.py and issue-config.sh

Then assert all 9 verbs resolve and --state blocked works.

Out of scope, noted

.agent/workflows/fix.md.backup and fix-loop.md.backup are stale editor backups committed to git.

🤖 Found by autonomous fix workflow

[laird]

PR #74 was blocked (CONFLICTING). Opened #87 against feature/issue-71 to clear it without force-pushing over #74's history.

The conflict was semantic: #74 replaces the stale mirror with the thin 9-verb dispatcher, while master has since added jira/ado backends. Resolving "ours" would have silently dropped Jira and Azure DevOps support from the .agent mirror. Resolution takes the canonical plugins file, which is both — and test_agent_mirror_parity.sh (added by #74) pins it as byte-identical, so it self-verifies.

Parity guard 13/13, 18/18 suites pass, build clean.

Merge order: #87 → feature/issue-71, then #74 → master.

🤖 Autonomous fix workflow