feat(plan-01): Phase 1 — relocate harness code to packages/casan-harness (symlink facade)

Physically move the pure-code subtrees out of .specify into the package, leaving compat symlinks at the old .specify/<dir> paths so every existing reference (internal CASAN_HARNESS_ROOT + external CI/docker/mjs) keeps resolving. Runtime state stays put. Moved (git mv): scripts/ tests/ security/ templates/ config/ governance/ memory/ .specify/<dir> -> packages/casan-harness/<dir> (+ .specify/<dir> symlink) Stays in .specify (state/governance/domain, handled later): logs/ agentops/ level5/ init-options.json traceability-map.json Python `.resolve()` self-location followed the compat symlink into packages and lost the app root; generate-casan-demo-context.py, generate-agentops-dashboard.py and dashboard-server.py now walk UP for the `.specify` state marker instead of a fixed parent depth (fixes "missing trace files" in run-casan4). Full gate: PASS=64 FAIL=0 SKIP=3 (CASAN_CI_STEP_TIMEOUT_SEC=1200 — track-a ~450s runs close to the 600s default and can tip over under load; this is timing variance, not a regression — it passed cleanly with headroom). Runtime log/audit artifacts kept unstaged. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-07-08 00:06:00 +09:00
parent 2c765c9a45
commit 664bd1f00c
229 changed files with 268 additions and 3 deletions
@@ -0,0 +1,52 @@
+#!/usr/bin/env bash
+set -uo pipefail
+
+# CASAN H4 — Tool-output indirect-injection scanner (Track A, V7).
+#
+# A tool (shell command, file read, web fetch, sub-agent) can return content
+# that is then fed back into a downstream model's context. If that content
+# carries a prompt injection, the model can be hijacked even though the ORIGINAL
+# user input was clean. This scans a tool-output file the same way an untrusted
+# artifact is scanned, BEFORE the output is allowed to re-enter model context.
+#
+# It reuses security-check.sh in `input` mode (block-pattern + unicode/encoding
+# normalization + secret detection) but forces the semantic/strict model path
+# OFF so the scan is deterministic and needs no model backend — this is a
+# pattern scan of machine output, not a user-intent classification.
+#
+# Usage:
+#   tool-output-scan.sh <tool-output-file> [context-label]
+# Exit:
+#   0  — safe to reuse
+#   2  — injection / secret pattern detected (caller should reject/quarantine)
+#   64 — usage error (file missing)
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+OUTPUT_FILE="${1:-}"
+LABEL="${2:-unknown-tool}"
+
+if [[ -z "$OUTPUT_FILE" || ! -f "$OUTPUT_FILE" ]]; then
+  echo "Usage: tool-output-scan.sh <tool-output-file> [context-label]" >&2
+  exit 64
+fi
+
+WORK="$(mktemp -d)"; trap 'rm -rf "$WORK"' EXIT
+SCAN_OUT="$WORK/tool-output-scan.txt"
+TIMESTAMP="$(date -u +"%Y-%m-%dT%H:%M:%SZ")"
+
+# Deterministic pattern scan: semantic + strict explicitly disabled here so a
+# tool-output scan never depends on (or is blocked by) model availability.
+CASAN_SECURITY_STRICT=0 CASAN_SEMANTIC_CLASSIFY=0 \
+  bash "$SCRIPT_DIR/security-check.sh" "$OUTPUT_FILE" "$SCAN_OUT" input >/dev/null 2>&1
+SC_RC=$?
+
+if [[ "$SC_RC" -eq 2 ]]; then
+  echo "TOOL_OUTPUT_SCAN_BLOCKED label=$LABEL file=$OUTPUT_FILE reason=injection_or_secret timestamp=$TIMESTAMP"
+  exit 2
+elif [[ "$SC_RC" -ne 0 ]]; then
+  echo "TOOL_OUTPUT_SCAN_ERROR label=$LABEL rc=$SC_RC" >&2
+  exit 2  # fail closed on scan error
+fi
+
+echo "TOOL_OUTPUT_SCAN_CLEAN label=$LABEL file=$OUTPUT_FILE timestamp=$TIMESTAMP"
+exit 0