crewly 1.15.0 → 1.17.0

package/config/skills/agent/core/gmail-send/execute.sh

@@ -0,0 +1,114 @@
+#!/bin/bash
+# =============================================================================
+# gmail-send — Send an email from the owner's Gmail
+#
+# Backed by POST /api/google/gmail/send. Refuses without --to and --subject.
+# With --dry-run or CREWLY_GMAIL_SEND_DRY_RUN=1 nothing is sent and no
+# request is made: the message preview is printed instead.
+#
+# Usage:
+#   bash execute.sh --to a@b.c --subject "Hi" --text "Hello" [--cc x@y.z]
+#                   [--thread-id <id>] [--in-reply-to "<msgid>"] [--dry-run]
+#   bash execute.sh '{"to":"a@b.c","subject":"Hi","text":"Hello"}'
+# =============================================================================
+set -euo pipefail
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+source "${SCRIPT_DIR}/../../_common/lib.sh"
+
+print_usage() {
+  cat <<'EOF_USAGE'
+Usage:
+  bash execute.sh --to a@b.c --subject "Hi" --text "Hello" [--cc x@y.z]
+                  [--thread-id <gmail thread id>] [--in-reply-to "<message-id>"] [--dry-run]
+  bash execute.sh '{"to":"a@b.c","subject":"Hi","text":"Hello"}'
+
+Options:
+  --to            Recipient(s), comma-separated (required)
+  --subject       Subject (required)
+  --text          Plain-text body (or --text-file <path>)
+  --text-file     Read the body from a file
+  --cc            Cc recipient(s)
+  --thread-id     Reply inside this Gmail thread
+  --in-reply-to   Message-ID being answered (sets In-Reply-To / References)
+  --dry-run       Preview only; nothing is sent (also: CREWLY_GMAIL_SEND_DRY_RUN=1)
+  --help | -h     Show this help
+EOF_USAGE
+}
+
+INPUT_JSON=""
+TO=""; CC=""; SUBJECT=""; TEXT=""; TEXT_FILE=""; THREAD_ID=""; IN_REPLY_TO=""
+DRY_RUN="${CREWLY_GMAIL_SEND_DRY_RUN:-}"
+
+if [[ $# -gt 0 && ${1:0:1} == '{' ]]; then
+  INPUT_JSON="$1"
+  shift || true
+fi
+
+while [[ $# -gt 0 ]]; do
+  case "$1" in
+    --to)          [ $# -ge 2 ] || error_exit "--to requires a value";          TO="$2";          shift 2 ;;
+    --cc)          [ $# -ge 2 ] || error_exit "--cc requires a value";          CC="$2";          shift 2 ;;
+    --subject)     [ $# -ge 2 ] || error_exit "--subject requires a value";     SUBJECT="$2";     shift 2 ;;
+    --text|--body) [ $# -ge 2 ] || error_exit "--text requires a value";        TEXT="$2";        shift 2 ;;
+    --text-file)   [ $# -ge 2 ] || error_exit "--text-file requires a value";   TEXT_FILE="$2";   shift 2 ;;
+    --thread-id)   [ $# -ge 2 ] || error_exit "--thread-id requires a value";   THREAD_ID="$2";   shift 2 ;;
+    --in-reply-to) [ $# -ge 2 ] || error_exit "--in-reply-to requires a value"; IN_REPLY_TO="$2"; shift 2 ;;
+    --dry-run)     DRY_RUN=1; shift ;;
+    --help|-h)     print_usage; exit 0 ;;
+    *) error_exit "Unknown option: $1" ;;
+  esac
+done
+
+if [ -n "$INPUT_JSON" ]; then
+  INPUT=$(read_json_input "$INPUT_JSON")
+  [ -z "$TO" ]          && TO=$(printf '%s' "$INPUT" | jq -r '.to // empty')
+  [ -z "$CC" ]          && CC=$(printf '%s' "$INPUT" | jq -r '.cc // empty')
+  [ -z "$SUBJECT" ]     && SUBJECT=$(printf '%s' "$INPUT" | jq -r '.subject // empty')
+  [ -z "$TEXT" ]        && TEXT=$(printf '%s' "$INPUT" | jq -r '.text // .body // empty')
+  [ -z "$THREAD_ID" ]   && THREAD_ID=$(printf '%s' "$INPUT" | jq -r '.threadId // empty')
+  [ -z "$IN_REPLY_TO" ] && IN_REPLY_TO=$(printf '%s' "$INPUT" | jq -r '.inReplyTo // empty')
+  [ -z "$DRY_RUN" ]     && DRY_RUN=$(printf '%s' "$INPUT" | jq -r 'if .dryRun == true then "1" else "" end')
+fi
+
+if [ -n "$TEXT_FILE" ]; then
+  [ -f "$TEXT_FILE" ] || error_exit "text file not found: $TEXT_FILE"
+  TEXT=$(cat "$TEXT_FILE")
+fi
+
+require_param "to (--to)" "$TO"
+require_param "subject (--subject)" "$SUBJECT"
+require_param "text (--text or --text-file)" "$TEXT"
+
+# Dry run: no request leaves this machine. Print a readable preview of the
+# headers and body exactly as they will be sent.
+if [ "$DRY_RUN" = "1" ]; then
+  PREVIEW="To: ${TO}"
+  [ -n "$CC" ] && PREVIEW="${PREVIEW}
+Cc: ${CC}"
+  PREVIEW="${PREVIEW}
+Subject: ${SUBJECT}"
+  [ -n "$IN_REPLY_TO" ] && PREVIEW="${PREVIEW}
+In-Reply-To: ${IN_REPLY_TO}"
+  [ -n "$THREAD_ID" ] && PREVIEW="${PREVIEW}
+X-Gmail-Thread: ${THREAD_ID}"
+  PREVIEW="${PREVIEW}
+
+${TEXT}"
+  jq -cn --arg p "$PREVIEW" '{success: true, dryRun: true, preview: $p}'
+  exit 0
+fi
+
+BODY=$(jq -cn --arg to "$TO" --arg cc "$CC" --arg subject "$SUBJECT" --arg text "$TEXT" \
+  --arg threadId "$THREAD_ID" --arg inReplyTo "$IN_REPLY_TO" \
+  '{to: $to, subject: $subject, text: $text}
+   + (if $cc != "" then {cc: $cc} else {} end)
+   + (if $threadId != "" then {threadId: $threadId} else {} end)
+   + (if $inReplyTo != "" then {inReplyTo: $inReplyTo} else {} end)')
+
+RESPONSE=$(api_call POST "/google/gmail/send" "$BODY" 2>&1) || {
+  printf '%s' "$RESPONSE" | jq -c '{success: false, reason: (.details.error // .details // .error // "unknown"), hint: (.details.hint // ""), message: (.details.message // "")}' 2>/dev/null \
+    || jq -n --arg r "$RESPONSE" '{success: false, reason: $r}'
+  exit 1
+}
+
+printf '%s' "$RESPONSE" | jq -c '{success: (.success // false), id: .data.id, threadId: .data.threadId}'

package/config/skills/agent/core/gmail-send/execute.test.sh

@@ -0,0 +1,83 @@
+#!/bin/bash
+# Tests for gmail-send — run with: bash execute.test.sh
+# Spins up a python HTTP stub as the backend, asserts the request body the
+# skill sends, the JSON it prints, and that dry-run makes NO request.
+set -u
+HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+EXEC="$HERE/execute.sh"
+PASS=0; FAIL=0
+check() {
+  local name="$1"; local got="$2"; local want="$3"
+  if [ "$got" = "$want" ]; then PASS=$((PASS+1)); else FAIL=$((FAIL+1)); echo "FAIL: $name"; echo "  want: $want"; echo "  got:  $got"; fi
+}
+
+PORT=18803
+STUB_LOG="$(mktemp)"; export STUB_LOG; export STUB_PORT="$PORT"
+python3 -u <<'PY' >/dev/null 2>&1 &
+from http.server import BaseHTTPRequestHandler, HTTPServer
+import json, os
+LOG = os.environ['STUB_LOG']; PORT = int(os.environ['STUB_PORT'])
+class H(BaseHTTPRequestHandler):
+    def do_POST(self):
+        n = int(self.headers.get('content-length', '0')); body = self.rfile.read(n).decode() if n else ''
+        data = json.loads(body or '{}')
+        open(LOG, 'w').write(json.dumps({'method': 'POST', 'path': self.path, 'body': data}))
+        if data.get('to') == 'no-grant@example.com':
+            self.send_response(409); self.send_header('Content-Type', 'application/json'); self.end_headers()
+            self.wfile.write(json.dumps({'success': False, 'error': 'not_connected', 'message': 'no grant', 'hint': 'https://cloud/start?token=j'}).encode()); return
+        self.send_response(200); self.send_header('Content-Type', 'application/json'); self.end_headers()
+        self.wfile.write(json.dumps({'success': True, 'data': {'id': 's1', 'threadId': data.get('threadId', 'tNew'), 'labelIds': ['SENT']}}).encode())
+    def log_message(self, *a, **k): pass
+HTTPServer(('127.0.0.1', PORT), H).serve_forever()
+PY
+STUB_PID=$!
+disown "$STUB_PID" 2>/dev/null || true
+trap 'kill "$STUB_PID" >/dev/null 2>&1; rm -f "$STUB_LOG"' EXIT
+for i in $(seq 1 30); do curl -s -o /dev/null -m 0.2 "http://127.0.0.1:${PORT}/" 2>/dev/null && break; sleep 0.1; done
+
+run() { CREWLY_API_URL="http://127.0.0.1:${PORT}" CREWLY_GMAIL_SEND_DRY_RUN="" bash "$EXEC" "$@" 2>/dev/null; }
+run_err() { CREWLY_API_URL="http://127.0.0.1:${PORT}" CREWLY_GMAIL_SEND_DRY_RUN="" bash "$EXEC" "$@" 2>&1 >/dev/null | jq -r '.error' 2>/dev/null; }
+
+# --- send with all fields ---
+OUT=$(run --to ann@example.com --cc bob@example.com --subject "Re: 第三季度" --text "Looks good" --thread-id t1 --in-reply-to "<abc@x>")
+check "send: output" "$OUT" '{"success":true,"id":"s1","threadId":"t1"}'
+check "send: path" "$(jq -r .path "$STUB_LOG")" "/api/google/gmail/send"
+check "send: body" "$(jq -c .body "$STUB_LOG")" '{"to":"ann@example.com","subject":"Re: 第三季度","text":"Looks good","cc":"bob@example.com","threadId":"t1","inReplyTo":"<abc@x>"}'
+
+# --- minimal + JSON input + text file ---
+TMP=$(mktemp); printf 'line one\nline two' > "$TMP"
+OUT=$(run '{"to":"a@b.c","subject":"Hi"}' --text-file "$TMP"); rm -f "$TMP"
+check "json+file: output" "$OUT" '{"success":true,"id":"s1","threadId":"tNew"}'
+check "json+file: body omits optional keys" "$(jq -c .body "$STUB_LOG")" '{"to":"a@b.c","subject":"Hi","text":"line one\nline two"}'
+
+# --- dry run via env: no request, preview printed ---
+: > "$STUB_LOG"
+OUT=$(CREWLY_API_URL="http://127.0.0.1:${PORT}" CREWLY_GMAIL_SEND_DRY_RUN=1 bash "$EXEC" --to a@b.c --cc c@d.e --subject "Hi" --text "hello" --in-reply-to "<m@x>" 2>/dev/null); RC=$?
+check "env dry-run: exit 0" "$RC" "0"
+check "env dry-run: preview" "$OUT" '{"success":true,"dryRun":true,"preview":"To: a@b.c\nCc: c@d.e\nSubject: Hi\nIn-Reply-To: <m@x>\n\nhello"}'
+check "env dry-run: no request made" "$(cat "$STUB_LOG")" ""
+
+# --- dry run via flag ---
+OUT=$(run --to a@b.c --subject "Hi" --text "x" --dry-run)
+check "flag dry-run: preview" "$OUT" '{"success":true,"dryRun":true,"preview":"To: a@b.c\nSubject: Hi\n\nx"}'
+check "flag dry-run: no request made" "$(cat "$STUB_LOG")" ""
+
+# --- dry run still validates ---
+check "dry-run without --to refuses" "$(CREWLY_GMAIL_SEND_DRY_RUN=1 run_err --subject Hi --text x)" "Missing required parameter: to (--to)"
+
+# --- backend 409 ---
+OUT=$(CREWLY_API_URL="http://127.0.0.1:${PORT}" CREWLY_GMAIL_SEND_DRY_RUN="" bash "$EXEC" --to no-grant@example.com --subject Hi --text x 2>/dev/null); RC=$?
+check "409: exit code" "$RC" "1"
+check "409: reason + hint" "$OUT" '{"success":false,"reason":"not_connected","hint":"https://cloud/start?token=j","message":"no grant"}'
+
+# --- validation (nothing sent) ---
+: > "$STUB_LOG"
+check "missing to errors" "$(run_err --subject Hi --text x)" "Missing required parameter: to (--to)"
+check "missing subject errors" "$(run_err --to a@b.c --text x)" "Missing required parameter: subject (--subject)"
+check "missing text errors" "$(run_err --to a@b.c --subject Hi)" "Missing required parameter: text (--text or --text-file)"
+check "missing text file errors" "$(run_err --to a@b.c --subject Hi --text-file /nonexistent/x.txt)" "text file not found: /nonexistent/x.txt"
+check "validation: no request made" "$(cat "$STUB_LOG")" ""
+check "unknown option errors" "$(run_err --bogus 1)" "Unknown option: --bogus"
+
+echo "=== Results: $PASS passed, $FAIL failed ==="
+[ "$FAIL" -eq 0 ]

package/config/skills/agent/core/record-prediction/SKILL.md

@@ -0,0 +1,61 @@
+---
+name: Record Prediction
+description: Record a prediction with a confidence level so your calibration can be measured once it resolves.
+version: 1.0.0
+category: memory
+skillType: claude-skill
+assignableRoles:
+  - developer
+  - qa
+  - tpm
+  - designer
+  - frontend-developer
+  - backend-developer
+  - fullstack-dev
+  - qa-engineer
+  - product-manager
+  - architect
+  - generalist
+  - sales
+  - support
+  - orchestrator
+triggers:
+  - record prediction
+  - I predict
+  - confidence
+  - estimate
+tags:
+  - memory
+  - calibration
+  - self-improvement
+execution:
+  type: script
+  script:
+    file: execute.sh
+    interpreter: bash
+    timeoutMs: 15000
+---
+
+# Record Prediction
+
+Write down a falsifiable prediction ("this PR lands today", "the flaky test is a race in the mock") with a confidence between 0 and 1. Later, close it with `resolve-prediction`. Your calibration score (how well confidence tracks reality) is shown in your prompt once predictions resolve, with guidance when you are over- or under-confident.
+
+## Parameters
+
+| Flag | JSON Field | Required | Description |
+|------|-----------|----------|-------------|
+| `--statement` / `-p` | `statement` | Yes | What you predict will happen |
+| `--confidence` / `-c` | `confidence` | Yes | 0–1 (e.g. `0.7`) |
+| `--resolve-by` | `resolveBy` | No | ISO date by which it should be resolved |
+| `--session` / `-s` | `sessionName` | No | Your session name — defaults to `$CREWLY_SESSION_NAME` |
+
+## Examples
+
+```bash
+bash execute.sh --statement "checkout PR merges before Friday" --confidence 0.7 --resolve-by 2026-10-03
+bash execute.sh '{"statement":"checkout PR merges before Friday","confidence":0.7}'
+```
+
+## Output
+
+`{ "success": true, "id": "pred-...", "prediction": {...} }` — keep the `id`; you need it to resolve.

package/config/skills/agent/core/record-prediction/execute.sh

@@ -0,0 +1,90 @@
+#!/bin/bash
+# =============================================================================
+# record-prediction — Record a prediction with a confidence level
+#
+# Backed by POST /api/agents/:sessionName/self-improvement/predictions.
+#
+# Usage:
+#   bash execute.sh --statement "PR lands today" --confidence 0.7 [--resolve-by 2026-10-01] [--session crewly-dev-1]
+#   bash execute.sh '{"statement":"PR lands today","confidence":0.7,"resolveBy":"2026-10-01"}'
+# =============================================================================
+set -euo pipefail
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+source "${SCRIPT_DIR}/../../_common/lib.sh"
+
+print_usage() {
+  cat <<'EOF_USAGE'
+Usage:
+  bash execute.sh --statement "PR lands today" --confidence 0.7 [--resolve-by 2026-10-01] [--session crewly-dev-1]
+  bash execute.sh '{"statement":"PR lands today","confidence":0.7}'
+
+Options:
+  --statement  | -p   What you predict (required)
+  --confidence | -c   Confidence 0-1 (required)
+  --resolve-by        ISO date the prediction should be resolved by (optional)
+  --session    | -s   Session name (defaults to $CREWLY_SESSION_NAME)
+  --help       | -h   Show this help
+EOF_USAGE
+}
+
+INPUT_JSON=""
+SESSION_NAME=""
+STATEMENT=""
+CONFIDENCE=""
+RESOLVE_BY=""
+
+if [[ $# -gt 0 && ${1:0:1} == '{' ]]; then
+  INPUT_JSON="$1"
+  shift || true
+fi
+
+while [[ $# -gt 0 ]]; do
+  case "$1" in
+    --statement|-p)
+      [ $# -ge 2 ] || error_exit "--statement requires a value"
+      STATEMENT="$2"; shift 2 ;;
+    --confidence|-c)
+      [ $# -ge 2 ] || error_exit "--confidence requires a value"
+      CONFIDENCE="$2"; shift 2 ;;
+    --resolve-by)
+      [ $# -ge 2 ] || error_exit "--resolve-by requires a value"
+      RESOLVE_BY="$2"; shift 2 ;;
+    --session|-s)
+      [ $# -ge 2 ] || error_exit "--session requires a value"
+      SESSION_NAME="$2"; shift 2 ;;
+    --help|-h) print_usage; exit 0 ;;
+    *) error_exit "Unknown option: $1" ;;
+  esac
+done
+
+if [ -n "$INPUT_JSON" ]; then
+  INPUT=$(read_json_input "$INPUT_JSON")
+  [ -z "$SESSION_NAME" ] && SESSION_NAME=$(printf '%s' "$INPUT" | jq -r '.sessionName // .agentId // empty')
+  [ -z "$STATEMENT" ] && STATEMENT=$(printf '%s' "$INPUT" | jq -r '.statement // .prediction // empty')
+  [ -z "$CONFIDENCE" ] && CONFIDENCE=$(printf '%s' "$INPUT" | jq -r 'if .confidence == null then empty else (.confidence | tostring) end')
+  [ -z "$RESOLVE_BY" ] && RESOLVE_BY=$(printf '%s' "$INPUT" | jq -r '.resolveBy // empty')
+fi
+
+[ -z "$SESSION_NAME" ] && SESSION_NAME="${CREWLY_SESSION_NAME:-}"
+require_param "sessionName (--session or CREWLY_SESSION_NAME)" "$SESSION_NAME"
+require_param "statement (--statement)" "$STATEMENT"
+require_param "confidence (--confidence)" "$CONFIDENCE"
+
+# Validate locally so a typo fails fast with a readable message.
+if ! printf '%s' "$CONFIDENCE" | grep -Eq '^(0(\.[0-9]+)?|1(\.0+)?)$'; then
+  error_exit "confidence must be a number between 0 and 1 (got: $CONFIDENCE)"
+fi
+
+if [ -n "$RESOLVE_BY" ]; then
+  BODY=$(jq -cn --arg s "$STATEMENT" --argjson c "$CONFIDENCE" --arg r "$RESOLVE_BY" '{statement: $s, confidence: $c, resolveBy: $r}')
+else
+  BODY=$(jq -cn --arg s "$STATEMENT" --argjson c "$CONFIDENCE" '{statement: $s, confidence: $c}')
+fi
+
+RESPONSE=$(api_call POST "/agents/${SESSION_NAME}/self-improvement/predictions" "$BODY" 2>&1) || {
+  ERROR_MSG=$(printf '%s' "$RESPONSE" | jq -r '.details.error // .details // .error // "Unknown error"' 2>/dev/null || echo "$RESPONSE")
+  jq -n --arg reason "$ERROR_MSG" '{success: false, reason: $reason}'
+  exit 1
+}
+
+printf '%s' "$RESPONSE" | jq -c '{success: (.success // false), id: (.data.prediction.id // null), prediction: (.data.prediction // null)}'

package/config/skills/agent/core/record-prediction/execute.test.sh

@@ -0,0 +1,67 @@
+#!/bin/bash
+# Tests for record-prediction — run with: bash execute.test.sh
+# Spins up a python HTTP stub as the backend, asserts the request path/body
+# the skill sends and the JSON it prints. Exit 0 on pass.
+set -u
+HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+EXEC="$HERE/execute.sh"
+PASS=0; FAIL=0
+check() {
+  local name="$1"; local got="$2"; local want="$3"
+  if [ "$got" = "$want" ]; then PASS=$((PASS+1)); else FAIL=$((FAIL+1)); echo "FAIL: $name"; echo "  want: $want"; echo "  got:  $got"; fi
+}
+
+PORT=18793
+STUB_LOG="$(mktemp)"; export STUB_LOG; export STUB_PORT="$PORT"
+python3 -u <<'PY' >/dev/null 2>&1 &
+from http.server import BaseHTTPRequestHandler, HTTPServer
+import json, os
+LOG = os.environ['STUB_LOG']; PORT = int(os.environ['STUB_PORT'])
+class H(BaseHTTPRequestHandler):
+    def do_POST(self):
+        n = int(self.headers.get('content-length', '0')); body = self.rfile.read(n).decode() if n else ''
+        data = json.loads(body or '{}')
+        open(LOG, 'w').write(json.dumps({'path': self.path, 'body': data}))
+        if data.get('resolveBy') == 'soon':
+            self.send_response(400); self.send_header('Content-Type', 'application/json'); self.end_headers()
+            self.wfile.write(b'{"success":false,"error":"resolveBy must be an ISO date string"}'); return
+        pred = {'id': 'pred-1-abc', 'prediction': data.get('statement'), 'confidence': data.get('confidence'), 'madeAt': '2026-09-18'}
+        if 'resolveBy' in data: pred['resolveBy'] = data['resolveBy']
+        self.send_response(201); self.send_header('Content-Type', 'application/json'); self.end_headers()
+        self.wfile.write(json.dumps({'success': True, 'data': {'prediction': pred}}).encode())
+    def log_message(self, *a, **k): pass
+HTTPServer(('127.0.0.1', PORT), H).serve_forever()
+PY
+STUB_PID=$!
+disown "$STUB_PID" 2>/dev/null || true
+trap 'kill "$STUB_PID" >/dev/null 2>&1; rm -f "$STUB_LOG"' EXIT
+for i in $(seq 1 30); do curl -s -o /dev/null -m 0.2 "http://127.0.0.1:${PORT}/" 2>/dev/null && break; sleep 0.1; done
+
+run() { CREWLY_API_URL="http://127.0.0.1:${PORT}" CREWLY_SESSION_NAME="crewly-dev-1" bash "$EXEC" "$@" 2>/dev/null; }
+run_err() { CREWLY_API_URL="http://127.0.0.1:${PORT}" CREWLY_SESSION_NAME="" bash "$EXEC" "$@" 2>&1 >/dev/null | jq -r '.error' 2>/dev/null; }
+
+OUT=$(run --statement "PR lands today" --confidence 0.7 --resolve-by 2026-10-01)
+check "flags: output id" "$(printf '%s' "$OUT" | jq -r .id)" "pred-1-abc"
+check "flags: output success" "$(printf '%s' "$OUT" | jq -r .success)" "true"
+check "flags: path" "$(jq -r .path "$STUB_LOG")" "/api/agents/crewly-dev-1/self-improvement/predictions"
+check "flags: body (confidence is numeric, resolveBy present)" "$(jq -c .body "$STUB_LOG")" '{"statement":"PR lands today","confidence":0.7,"resolveBy":"2026-10-01"}'
+
+run '{"statement":"no deadline","confidence":1,"sessionName":"crewly-qa-9"}' >/dev/null
+check "json: body omits resolveBy when absent" "$(jq -c .body "$STUB_LOG")" '{"statement":"no deadline","confidence":1}'
+check "json: explicit session in path" "$(jq -r .path "$STUB_LOG")" "/api/agents/crewly-qa-9/self-improvement/predictions"
+
+run '{"prediction":"legacy field name","confidence":"0.25"}' >/dev/null
+check "json: accepts legacy 'prediction' key and string confidence" "$(jq -c .body "$STUB_LOG")" '{"statement":"legacy field name","confidence":0.25}'
+
+OUT=$(CREWLY_API_URL="http://127.0.0.1:${PORT}" CREWLY_SESSION_NAME="crewly-dev-1" bash "$EXEC" --statement x --confidence 0.5 --resolve-by soon 2>/dev/null); RC=$?
+check "backend 400: exit code" "$RC" "1"
+check "backend 400: reason" "$(printf '%s' "$OUT" | jq -r .reason)" "resolveBy must be an ISO date string"
+
+check "confidence > 1 rejected locally" "$(run_err --session s --statement x --confidence 1.5)" "confidence must be a number between 0 and 1 (got: 1.5)"
+check "non-numeric confidence rejected locally" "$(run_err --session s --statement x --confidence abc)" "confidence must be a number between 0 and 1 (got: abc)"
+check "missing statement errors" "$(run_err --session s --confidence 0.5)" "Missing required parameter: statement (--statement)"
+check "missing confidence errors" "$(run_err --session s --statement x)" "Missing required parameter: confidence (--confidence)"
+check "missing session errors" "$(run_err --statement x --confidence 0.5)" "Missing required parameter: sessionName (--session or CREWLY_SESSION_NAME)"
+
+echo "=== Results: $PASS passed, $FAIL failed ==="
+[ "$FAIL" -eq 0 ]

package/config/skills/agent/core/resolve-prediction/SKILL.md

@@ -0,0 +1,61 @@
+---
+name: Resolve Prediction
+description: Close a recorded prediction with what actually happened so your calibration score updates.
+version: 1.0.0
+category: memory
+skillType: claude-skill
+assignableRoles:
+  - developer
+  - qa
+  - tpm
+  - designer
+  - frontend-developer
+  - backend-developer
+  - fullstack-dev
+  - qa-engineer
+  - product-manager
+  - architect
+  - generalist
+  - sales
+  - support
+  - orchestrator
+triggers:
+  - resolve prediction
+  - prediction outcome
+  - was I right
+tags:
+  - memory
+  - calibration
+  - self-improvement
+execution:
+  type: script
+  script:
+    file: execute.sh
+    interpreter: bash
+    timeoutMs: 15000
+---
+
+# Resolve Prediction
+
+Record the outcome of a prediction made with `record-prediction`. Pass `--accurate true|false`; if you only pass an outcome, a plain verdict such as `correct`, `wrong`, `yes`, `no` is understood — anything else needs the explicit flag. The response includes your updated calibration score.
+
+## Parameters
+
+| Flag | JSON Field | Required | Description |
+|------|-----------|----------|-------------|
+| `--id` | `id` | Yes | Prediction id (`pred-...`) |
+| `--outcome` / `-o` | `outcome` | Yes | What actually happened |
+| `--accurate` / `-a` | `accurate` | No* | `true` / `false` — required unless outcome is a plain verdict |
+| `--session` / `-s` | `sessionName` | No | Your session name — defaults to `$CREWLY_SESSION_NAME` |
+
+## Examples
+
+```bash
+bash execute.sh --id pred-1718000000-ab12cd --outcome "merged Thursday" --accurate true
+bash execute.sh --id pred-1718000000-ab12cd --outcome wrong
+bash execute.sh '{"id":"pred-1718000000-ab12cd","outcome":"slipped a week","accurate":false}'
+```
+
+## Output
+
+`{ "success": true, "prediction": {...}, "calibrationScore": 0.82 }`

package/config/skills/agent/core/resolve-prediction/execute.sh

@@ -0,0 +1,87 @@
+#!/bin/bash
+# =============================================================================
+# resolve-prediction — Close a prediction with its actual outcome
+#
+# Backed by POST /api/agents/:sessionName/self-improvement/predictions/:id/resolve.
+#
+# Usage:
+#   bash execute.sh --id pred-123 --outcome "merged Thursday" --accurate true [--session crewly-dev-1]
+#   bash execute.sh --id pred-123 --outcome wrong
+#   bash execute.sh '{"id":"pred-123","outcome":"slipped","accurate":false}'
+# =============================================================================
+set -euo pipefail
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+source "${SCRIPT_DIR}/../../_common/lib.sh"
+
+print_usage() {
+  cat <<'EOF_USAGE'
+Usage:
+  bash execute.sh --id pred-123 --outcome "merged Thursday" --accurate true [--session crewly-dev-1]
+  bash execute.sh --id pred-123 --outcome wrong
+  bash execute.sh '{"id":"pred-123","outcome":"slipped","accurate":false}'
+
+Options:
+  --id               Prediction id (required)
+  --outcome  | -o    What actually happened (required)
+  --accurate | -a    true|false (required unless outcome is a plain verdict like "correct"/"wrong")
+  --session  | -s    Session name (defaults to $CREWLY_SESSION_NAME)
+  --help     | -h    Show this help
+EOF_USAGE
+}
+
+INPUT_JSON=""
+SESSION_NAME=""
+PRED_ID=""
+OUTCOME=""
+ACCURATE=""
+
+if [[ $# -gt 0 && ${1:0:1} == '{' ]]; then
+  INPUT_JSON="$1"
+  shift || true
+fi
+
+while [[ $# -gt 0 ]]; do
+  case "$1" in
+    --id)
+      [ $# -ge 2 ] || error_exit "--id requires a value"
+      PRED_ID="$2"; shift 2 ;;
+    --outcome|-o)
+      [ $# -ge 2 ] || error_exit "--outcome requires a value"
+      OUTCOME="$2"; shift 2 ;;
+    --accurate|-a)
+      [ $# -ge 2 ] || error_exit "--accurate requires a value"
+      ACCURATE="$2"; shift 2 ;;
+    --session|-s)
+      [ $# -ge 2 ] || error_exit "--session requires a value"
+      SESSION_NAME="$2"; shift 2 ;;
+    --help|-h) print_usage; exit 0 ;;
+    *) error_exit "Unknown option: $1" ;;
+  esac
+done
+
+if [ -n "$INPUT_JSON" ]; then
+  INPUT=$(read_json_input "$INPUT_JSON")
+  [ -z "$SESSION_NAME" ] && SESSION_NAME=$(printf '%s' "$INPUT" | jq -r '.sessionName // .agentId // empty')
+  [ -z "$PRED_ID" ] && PRED_ID=$(printf '%s' "$INPUT" | jq -r '.id // .predictionId // empty')
+  [ -z "$OUTCOME" ] && OUTCOME=$(printf '%s' "$INPUT" | jq -r '.outcome // empty')
+  [ -z "$ACCURATE" ] && ACCURATE=$(printf '%s' "$INPUT" | jq -r 'if .accurate == null then empty else (.accurate | tostring) end')
+fi
+
+[ -z "$SESSION_NAME" ] && SESSION_NAME="${CREWLY_SESSION_NAME:-}"
+require_param "sessionName (--session or CREWLY_SESSION_NAME)" "$SESSION_NAME"
+require_param "id (--id)" "$PRED_ID"
+require_param "outcome (--outcome)" "$OUTCOME"
+
+case "$ACCURATE" in
+  "") BODY=$(jq -cn --arg o "$OUTCOME" '{outcome: $o}') ;;
+  true|false) BODY=$(jq -cn --arg o "$OUTCOME" --argjson a "$ACCURATE" '{outcome: $o, accurate: $a}') ;;
+  *) error_exit "accurate must be true or false (got: $ACCURATE)" ;;
+esac
+
+RESPONSE=$(api_call POST "/agents/${SESSION_NAME}/self-improvement/predictions/${PRED_ID}/resolve" "$BODY" 2>&1) || {
+  ERROR_MSG=$(printf '%s' "$RESPONSE" | jq -r '.details.error // .details // .error // "Unknown error"' 2>/dev/null || echo "$RESPONSE")
+  jq -n --arg reason "$ERROR_MSG" '{success: false, reason: $reason}'
+  exit 1
+}
+
+printf '%s' "$RESPONSE" | jq -c '{success: (.success // false), prediction: (.data.prediction // null), calibrationScore: (.data.calibrationScore // null)}'

package/config/skills/agent/core/resolve-prediction/execute.test.sh

@@ -0,0 +1,74 @@
+#!/bin/bash
+# Tests for resolve-prediction — run with: bash execute.test.sh
+# Spins up a python HTTP stub as the backend, asserts the request path/body
+# the skill sends and the JSON it prints. Exit 0 on pass.
+set -u
+HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+EXEC="$HERE/execute.sh"
+PASS=0; FAIL=0
+check() {
+  local name="$1"; local got="$2"; local want="$3"
+  if [ "$got" = "$want" ]; then PASS=$((PASS+1)); else FAIL=$((FAIL+1)); echo "FAIL: $name"; echo "  want: $want"; echo "  got:  $got"; fi
+}
+
+PORT=18794
+STUB_LOG="$(mktemp)"; export STUB_LOG; export STUB_PORT="$PORT"
+python3 -u <<'PY' >/dev/null 2>&1 &
+from http.server import BaseHTTPRequestHandler, HTTPServer
+import json, os
+LOG = os.environ['STUB_LOG']; PORT = int(os.environ['STUB_PORT'])
+class H(BaseHTTPRequestHandler):
+    def do_POST(self):
+        n = int(self.headers.get('content-length', '0')); body = self.rfile.read(n).decode() if n else ''
+        data = json.loads(body or '{}')
+        open(LOG, 'w').write(json.dumps({'path': self.path, 'body': data}))
+        self.send_header_json = lambda code: (self.send_response(code), self.send_header('Content-Type', 'application/json'), self.end_headers())
+        if '/pred-missing/' in self.path:
+            self.send_header_json(404); self.wfile.write(b'{"success":false,"error":"Prediction pred-missing not found"}'); return
+        if 'accurate' not in data and data.get('outcome') not in ('correct', 'wrong'):
+            self.send_header_json(400); self.wfile.write(b'{"success":false,"error":"accurate (boolean) is required"}'); return
+        acc = data.get('accurate', data.get('outcome') == 'correct')
+        pred = {'id': 'pred-1-abc', 'prediction': 'x', 'confidence': 0.9, 'madeAt': '2026-09-18', 'outcome': data.get('outcome'), 'accurate': acc, 'resolvedAt': '2026-09-18'}
+        self.send_header_json(200)
+        self.wfile.write(json.dumps({'success': True, 'data': {'prediction': pred, 'calibrationScore': 0.82}}).encode())
+    def log_message(self, *a, **k): pass
+HTTPServer(('127.0.0.1', PORT), H).serve_forever()
+PY
+STUB_PID=$!
+disown "$STUB_PID" 2>/dev/null || true
+trap 'kill "$STUB_PID" >/dev/null 2>&1; rm -f "$STUB_LOG"' EXIT
+for i in $(seq 1 30); do curl -s -o /dev/null -m 0.2 "http://127.0.0.1:${PORT}/" 2>/dev/null && break; sleep 0.1; done
+
+run() { CREWLY_API_URL="http://127.0.0.1:${PORT}" CREWLY_SESSION_NAME="crewly-dev-1" bash "$EXEC" "$@" 2>/dev/null; }
+run_err() { CREWLY_API_URL="http://127.0.0.1:${PORT}" CREWLY_SESSION_NAME="" bash "$EXEC" "$@" 2>&1 >/dev/null | jq -r '.error' 2>/dev/null; }
+
+OUT=$(run --id pred-1-abc --outcome "merged Thursday" --accurate true)
+check "flags: success + calibration" "$(printf '%s' "$OUT" | jq -c '{success, calibrationScore}')" '{"success":true,"calibrationScore":0.82}'
+check "flags: prediction accurate" "$(printf '%s' "$OUT" | jq -r .prediction.accurate)" "true"
+check "flags: path includes id" "$(jq -r .path "$STUB_LOG")" "/api/agents/crewly-dev-1/self-improvement/predictions/pred-1-abc/resolve"
+check "flags: body (accurate is boolean)" "$(jq -c .body "$STUB_LOG")" '{"outcome":"merged Thursday","accurate":true}'
+
+run --id pred-1-abc --outcome wrong >/dev/null
+check "verdict-only outcome omits accurate" "$(jq -c .body "$STUB_LOG")" '{"outcome":"wrong"}'
+
+run '{"id":"pred-1-abc","outcome":"slipped","accurate":false,"sessionName":"crewly-qa-9"}' >/dev/null
+check "json: body" "$(jq -c .body "$STUB_LOG")" '{"outcome":"slipped","accurate":false}'
+check "json: explicit session in path" "$(jq -r .path "$STUB_LOG")" "/api/agents/crewly-qa-9/self-improvement/predictions/pred-1-abc/resolve"
+
+run '{"predictionId":"pred-1-abc","outcome":"correct"}' >/dev/null
+check "json: accepts legacy predictionId key" "$(jq -r .path "$STUB_LOG")" "/api/agents/crewly-dev-1/self-improvement/predictions/pred-1-abc/resolve"
+
+OUT=$(CREWLY_API_URL="http://127.0.0.1:${PORT}" CREWLY_SESSION_NAME="crewly-dev-1" bash "$EXEC" --id pred-missing --outcome correct 2>/dev/null); RC=$?
+check "backend 404: exit code" "$RC" "1"
+check "backend 404: reason" "$(printf '%s' "$OUT" | jq -r .reason)" "Prediction pred-missing not found"
+
+OUT=$(CREWLY_API_URL="http://127.0.0.1:${PORT}" CREWLY_SESSION_NAME="crewly-dev-1" bash "$EXEC" --id pred-1-abc --outcome "took a while" 2>/dev/null); RC=$?
+check "backend 400 (ambiguous outcome): reason" "$(printf '%s' "$OUT" | jq -r .reason)" "accurate (boolean) is required"
+
+check "bad accurate value rejected locally" "$(run_err --session s --id p --outcome x --accurate maybe)" "accurate must be true or false (got: maybe)"
+check "missing id errors" "$(run_err --session s --outcome x)" "Missing required parameter: id (--id)"
+check "missing outcome errors" "$(run_err --session s --id p)" "Missing required parameter: outcome (--outcome)"
+check "missing session errors" "$(run_err --id p --outcome x)" "Missing required parameter: sessionName (--session or CREWLY_SESSION_NAME)"
+
+echo "=== Results: $PASS passed, $FAIL failed ==="
+[ "$FAIL" -eq 0 ]