@cxi-lmai/ci-agent-platform 3.1.1 → 3.1.2

package/README.md

@@ -144,6 +144,13 @@ picked, not both. Claude-harness model usage is billed by Anthropic, while
 omp-harness model usage is billed by OpenRouter. GitLab only for now; GitHub
 Actions omp support is not shipped yet.
 
+`omp` is a Bun program, so the GitLab template's `before_script` installs
+`bun` next to it (`npm install -g bun`) — the default `node:22-bookworm` image
+ships no Bun runtime, and without it every agent call fails while the job
+still reports success. A custom `PIPE_CI_IMAGE` needs nothing beyond node +
+npm for that install to work. If the harness binary cannot execute, the job
+now fails immediately instead of going green with no review.
+
 ### What it costs
 
 Every pipeline job spends paid model usage, so the defaults are deliberately

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@cxi-lmai/ci-agent-platform",
-  "version": "3.1.1",
+  "version": "3.1.2",
   "description": "Autonomous dev pipeline on plain GitLab CI or GitHub Actions, driven by Claude Code. A labeled issue goes in, an open merge request comes out.",
   "keywords": [
     "claude",

package/payload/ci-templates/claude-pipeline.gitlab-ci.yml

@@ -31,8 +31,9 @@ variables:
   PIPE_SCRIPTS_DIR: ".claude-pipeline/scripts"
 
   # --- CI image (see ci-templates decision note) -----------------------------
-  # Must provide: node + npm, git, curl. jq and claude-code are installed in
-  # before_script when missing. The runner scripts talk to the GitLab API with
+  # Must provide: node + npm, git, curl. jq, claude-code, and (when
+  # PIPE_HARNESS=omp) bun + omp are installed in before_script when missing.
+  # The runner scripts talk to the GitLab API with
   # curl and jq directly, so no glab is needed in the image; glab is only used
   # interactively, by the install wizard.
   # Python caveat (E2E finding N-7): node:22-bookworm ships python3 but NOT
@@ -162,7 +163,7 @@ variables:
 
 # --- Hidden base job: shared config + runtime variable mapping ---------------
 # Maps GitLab's CI_ runtime variables onto the PIPE_ names the scripts read, and
-# installs claude-code if the image does not ship it.
+# installs the selected harness (and its runtime) if the image does not ship it.
 .claude-base:
   image: $PIPE_CI_IMAGE
   variables:
@@ -193,6 +194,12 @@ variables:
       fi
     - |
       if [ "$PIPE_HARNESS" = "omp" ]; then
+        # omp is a Bun program (`#!/usr/bin/env bun`, engines.bun >= 1.3.14)
+        # and node:22-bookworm ships no bun, so the installed `omp` shim is
+        # unrunnable without it. The `bun` npm package postinstalls the real
+        # binary, so no curl|bash and no image change is needed. Deliberately
+        # no `|| true`: a missing harness runtime must fail the job loudly.
+        command -v bun >/dev/null 2>&1 || npm install -g bun
         command -v omp >/dev/null 2>&1 || npm install -g @oh-my-pi/pi-coding-agent
       else
         command -v claude >/dev/null 2>&1 || npm install -g @anthropic-ai/claude-code

package/payload/ci-templates/scripts/lib/pipeline-common.sh

@@ -246,12 +246,39 @@ pipe_scrub_agent_secrets() {
   done < <(compgen -e | while IFS= read -r v; do pipe_agent_secret_var "$v" && echo "$v"; done)
 }
 
+# --- Harness preflight ------------------------------------------------------
+# Every pipe_run_* invocation ends in `|| true`, so a harness binary that
+# exists but cannot execute used to produce a green job with no review, no
+# marker file and one line on stderr (issue #1: `omp` is a Bun program and the
+# default node:22-bookworm image ships no bun, so `command -v omp` succeeded
+# while every run failed with "/usr/bin/env: 'bun': No such file or
+# directory"). Executing the binary once, before the tolerant invocation,
+# converts any harness-runtime breakage into an immediate job failure.
+
+pipe_require_harness() {
+  # $1 = harness binary name (claude|omp). Exits the job on failure: a
+  # non-runnable harness means no agent can run, so there is nothing left for
+  # the caller to recover from.
+  local bin="$1" out
+  if ! command -v "$bin" >/dev/null 2>&1; then
+    pipe_log "FATAL: harness binary '$bin' not found on PATH (PIPE_HARNESS=${PIPE_HARNESS:-claude}); install it in before_script"
+    exit 1
+  fi
+  if ! out=$("$bin" --version 2>&1); then
+    pipe_log "FATAL: harness binary '$bin' ($(command -v "$bin")) cannot execute: '$bin --version' failed"
+    printf '%s\n' "$out" >&2
+    pipe_log "omp requires bun on PATH (npm install -g bun); claude requires node"
+    exit 1
+  fi
+}
+
 pipe_run_claude() {
   # $1 = agent label for metrics, $2 = skill invocation (e.g. "/review-mr"),
   # $3 = allowed tools (optional), $4 = model for the main loop (optional,
   # e.g. $PIPE_MODEL_REVIEW; subagents keep their frontmatter models),
   # stdin = extra prompt (optional, usually empty)
   local label="$1" skill="$2" tools="${3:-Agent,Read,Write,Edit,Glob,Grep,Bash}" model="${4:-}"
+  pipe_require_harness claude
   local stdin_file; stdin_file=$(mktemp)
   cat > "$stdin_file" || true
 
@@ -350,6 +377,7 @@ pipe_run_omp() {
   # frontmatter), stdin = extra prompt (optional, usually empty).
   local label="$1" skill="$2" tools="${3:-Agent,Read,Write,Edit,Glob,Grep,Bash}" model="${4:-}"
   local omp_tools; omp_tools=$(pipe_translate_tools_to_omp "$tools")
+  pipe_require_harness omp
   local stdin_file; stdin_file=$(mktemp)
   cat > "$stdin_file" || true