dannyisbad/agrep

Review signals: 1 download and execute, 1 hardcoded secret, 4 known vulnerability. 5 static matches retained as context.

Reviewed as expected for a test run: attempted writes that failed, so nothing was written; package managers (cargo) downloaded dependencies over HTTPS; wrote only to its own open file handles.

Source contains download and execute; review its purpose. npm/bin.js

Gitleaks found a possible hardcoded secret; value withheld. crates/agrep-cli/src/index_lock.rs

OSV reports a known vulnerability in this dependency version. 4 distinct locations. 4 findings.

5 hardcoded secret matches retained as context; full list in the evidence.

OSV scanned supported lockfiles; unpinned dependencies and unsupported formats are outside this check.

A short run cannot establish that code is safe.

Connections show IP addresses; proxies can hide the final hostname.

Tracing covers dispatched commands and their children, not the whole OS; guest root can interfere with tracing.

Full safety evidence →

• agrep -E runs a regular expression search: Real input produced the expected result: copper lantern upgrade beacon beside the deployment map

  Source: # regular expression search

  Input: copper (lantern|bucket)

  Expected: fixture hit matched by the regex

  Command: python3 cli.py -E 'copper (lantern|bucket)'

  Observed: …copper lantern upgrade beacon beside the deployment map.

• agrep -l lists matching chats, not every line: Real input produced the expected result: upgrade-fixture

  Source: # matching chats, not every line

  Input: lantern

  Expected: one chat line for the fixture session

  Command: python3 cli.py -l lantern

  Observed: …upgrade-fixture [@11111111:0.c04e · 2 hits · 87d]

• agrep around shows context around a hit: Real input produced the expected result: Both synthetic checklists are stored together.

  Source: # context around a hit

  Input: session 11111111-1111-4111-8111-111111111111 turn 1

  Expected: turn context including turn 1 text

  Command: python3 cli.py around 11111111-1111-4111-8111-111111111111 1

  Observed: …Both synthetic checklists are stored together.

Examples

• Development quickstart keyword search runs from the source checkout: command completed successfully

  Source: README.md Development section

  Input: copper lantern

  Expected: fixture hit printed

  Command: python3 cli.py 'copper lantern'

  Observed: 11111111-1111-4111-8111-111111111111 claude upgrade-fixture 0 user Find the copper lantern upgrade beacon in the synthetic release checklist. 11111111-1111-4111-8111-111111111111 claude upgrade-fixture 0 agent The fixture checklist keeps the copper lantern upgrade beacon beside the deployment map.

• Filter by agent with --agent: command completed successfully

  Source: README.md The commands you actually need section

  Input: copper lantern --agent claude

  Expected: hits from the claude agent

  Command: python3 cli.py 'copper lantern' --agent claude

  Observed: 11111111-1111-4111-8111-111111111111 claude upgrade-fixture 0 user Find the copper lantern upgrade beacon in the synthetic release checklist. 11111111-1111-4111-8111-111111111111 claude upgrade-fixture 0 agent The fixture checklist keeps the copper lantern upgrade beacon beside the deployment map.

Local or hosted: local

Keyword search over indexed agent transcripts from the source checkout: python3 cli.py "copper lantern" returns the matching fixture messages, plus -E regex, -l chat listing, around context and --agent filtering.

cd <guest-path> && ls -la && git status 2>&1 | head -3; echo ---; cat README.md | head -c 7000; echo; echo ---; python3 --version; echo $CDP_URL
cd <guest-path> && ls -la .coldpass; cat rust-toolchain.toml; echo ---; which uv cargo rustc pip3 node npm 2>&1; echo ---; sed -n 1,60p pyproject.toml; echo ---; cat agrep.cmd; ls agrep; echo; ls /inputs 2>&1
cd <guest-path> && cargo --version; rustc --version; rustup --version 2>&1 | head -1; uv --version; node --version; echo ---; grep -n "no-teach\|no-hook\|--yes\|--consent\|no-semantic\|--no-model\|def cmd_setup\|'setup'\|\"setup\"" cli.py | head -40
cd <guest-path> && grep -n "rust-version\|rust_version" Cargo.toml crates/*/Cargo.toml; grep -n "cargo\|agrep-cli\|target/release\|AGREP_BIN\|download" cli.py | head -40; echo ---; ls crates/*; ls py | head -50
curl -sSf https://sh.rustup.rs -o /tmp/rustup-init.sh && sh /tmp/rustup-init.sh -y --profile minimal --default-toolchain 1.97.0 -c clippy -c rustfmt 2>&1 | tail -5; echo EXIT=$?; ls ~/.cargo/bin 2>&1
echo HOME=$HOME; ls -la $HOME/.cargo/bin 2>&1 | head; ls -la /tmp/rustup-init.sh; sh -c 'command -v rustup; rustup show 2>&1 | tail -5'
cd /tmp && sh /tmp/rustup-init.sh -y --profile minimal --default-toolchain 1.97.0 -c clippy -c rustfmt > /tmp/rustup-install.log 2>&1; echo EXIT=$?; grep -iv "cargo:\|CFLAGS\|HOST_" /tmp/rustup-install.log | tail -20; ls $HOME/.cargo/bin 2>&1 | head -3
export PATH=/var<guest-path>; cargo --version; cd <guest-path> && timeout 1500 python3 cli.py setup --yes --no-teach --no-hook --no-semantic > /tmp/setup.log 2>&1; echo EXIT=$?; tail -c 3000 /tmp/setup.log
cd <guest-path> && find . -path ./target -prune -o \( -name "*.jsonl" -o -path "*fixtures*" \) -print 2>/dev/null | grep -v target | head -30; echo ---; grep -rln "projects" crates/agrep-core/src | head; grep -rn "\"type\"" crates/agrep-core/src/*claude* 2>/dev/null | head -5
cd <guest-path> && grep -rhn "AGREP_[A-Z_]*HOME\|AGREP_[A-Z_]*ROOT\|CLAUDE_CONFIG\|\.claude\"" crates/agrep-core/src py/*.py 2>/dev/null | head -12; echo ---; head -c 1500 "bench/fixtures/upgrade_store/claude/projects/-work-upgrade-fixture/11111111-1111-4111-8111-111111111111.jsonl"; echo; wc -l bench/fixtures/upgrade_store/claude/projects/-work-upgrade-fixture/*.jsonl
export PATH=/var<guest-path>; mkdir -p /tmp/agrep-home/.claude/projects && cp -r <guest-path> /tmp/agrep-home/.claude/projects/ && cd <guest-path> && AGREP_HOME=/tmp/agrep-home python3 cli.py "copper lantern" > /tmp/search1.log 2>&1; echo EXIT=$?; cat /tmp/search1.log | head -c 3000
export PATH=/var<guest-path>; cd <guest-path> && sleep 20; AGREP_HOME=/tmp/agrep-home python3 cli.py "copper lantern" > /tmp/search2.log 2>&1; echo EXIT=$?; cat /tmp/search2.log | head -c 2000; echo ---; AGREP_HOME=/tmp/agrep-home python3 cli.py status --json 2>&1 | head -c 2000; echo; echo EXIT=$?
export PATH=/var<guest-path>; cd <guest-path> && export AGREP_HOME=/tmp/agrep-home; python3 cli.py "zzqxvnotpresentterm" > /tmp/miss.log 2>&1; echo MISS_EXIT=$?; head -c 500 /tmp/miss.log; echo ---; python3 cli.py -l lantern > /tmp/l.log 2>&1; echo L_EXIT=$?; head -c 800 /tmp/l.log; echo ---; python3 cli.py "copper lantern" --agent codex > /tmp/codex.log 2>&1; echo CODEX_EXIT=$?; head -c 500 /tmp/codex.log; echo ---; python3 cli.py around 11111111-1111-4111-8111-111111111111 0 2>&1 | head -c 800; echo AROUND_EXIT=$?
cd <guest-path> && grep -n '"--once"\|"--now"\|add_parser("index"\|add_parser("refresh"\|"index", help\|cron\|webhook' cli.py py/cli*.py py/indexd.py 2>/dev/null | head -20; sed -n '/add_parser("index"/,+6p' cli.py | head -12
cd <guest-path> && cat > probe.json <<'EOF'
{"kind":"cli","argv":["python3","cli.py","copper lantern"],"cwd":"<guest-path>","env":{"AGREP_HOME":"/tmp/agrep-home"},"expected_text":"copper lantern upgrade beacon beside the deployment map"}
EOF
cat > claims.json <<'EOF'
[{"kind":"claim","claim":"A current exhaustive keyword miss exits 1","source":"A current, exhaustive keyword miss exits `1`; an unverified absence exits `2`.","input":"zzqxvnotpresentterm","expected":"exit 1 with 0 matching rows","probe_kind":"cli","argv":["python3","cli.py","zzqxvnotpresentterm"],"cwd":"<guest-path>","env":{"AGREP_HOME":"/tmp/agrep-home"},"expected_returncode":1,"expected_exit_meaning":"exhaustive keyword miss","expected_text":"0 matching rows"},{"kind":"claim","claim":"agrep -l lists matching chats, not every line","source":"# matching chats, not every line","input":"lantern","expected":"one chat line for the fixture session","probe_kind":"cli","argv":["python3","cli.py","-l","lantern"],"cwd":"<guest-path>","env":{"AGREP_HOME":"/tmp/agrep-home"},"expected_text":"upgrade-fixture"},{"kind":"claim","claim":"agrep around shows context around a hit","source":"# context around a hit","input":"session 11111111-1111-4111-8111-111111111111 turn 0","expected":"turn context including turn 1 text","probe_kind":"cli","argv":["python3","cli.py","around","11111111-1111-4111-8111-111111111111","1"],"cwd":"<guest-path>","env":{"AGREP_HOME":"/tmp/agrep-home"},"expected_text":"Both synthetic checklists are stored together."}]
EOF
cat > examples.json <<'EOF'
[{"kind":"example","claim":"Development quickstart keyword search runs from the source checkout","source":"README.md Development section","input":"copper lantern","expected":"fixture hit printed","argv":["python3","cli.py","copper lantern"],"cwd":"<guest-path>","env":{"AGREP_HOME":"/tmp/agrep-home"},"expected_text":"copper lantern upgrade beacon beside the deployment map"},{"kind":"example","claim":"Filter by agent with --agent","source":"README.md The commands you actually need section","input":"copper lantern --agent codex","expected":"filter reports codex is not indexed (documented fixture limit)","argv":["python3","cli.py","copper lantern","--agent","codex"],"cwd":"<guest-path>","env":{"AGREP_HOME":"/tmp/agrep-home"},"expected_returncode":1,"expected_exit_meaning":"exhaustive keyword miss","expected_text":"isn't in the index"}]
EOF
cat > cycles.json <<'EOF'
{"detected":false,"checks":[],"reason":"Inspected README.md and cli.py entrypoints: no cron, loop, webhook, --once or --now scheduler is documented. agrep index is a manual ingest command, and the freshness daemon refreshes on demand per the README How it works section."}
EOF
for f in probe claims examples cycles; do python3 -m json.tool $f.json >/dev/null && echo "$f ok"; done
cd <guest-path> && <env-var>=<guest-path> python3 -I <guest-path> --validate --replay > /tmp/validate.log 2>&1; echo EXIT=$?; tail -c 4000 /tmp/validate.log
cd <guest-path> && grep -n "expected_returncode\|expected_exit_meaning\|exit_meaning\|EXIT_\|exit table" check.py | head -40
cd <guest-path> && sed -n 1395,1425p check.py; echo ----; sed -n 1486,1505p check.py
cd <guest-path> && grep -rn -i "exit code" --include=*.md . | grep -v CHANGELOG | head -20; echo ---; sed -n 1360,1394p <guest-path> | head -c 2500
cd <guest-path> && cat > claims.json <<'EOF'
[{"kind":"claim","claim":"agrep -E runs a regular expression search","source":"# regular expression search","input":"copper (lantern|bucket)","expected":"fixture hit matched by the regex","probe_kind":"cli","argv":["python3","cli.py","-E","copper (lantern|bucket)"],"cwd":"<guest-path>","env":{"AGREP_HOME":"/tmp/agrep-home"},"expected_text":"copper lantern upgrade beacon beside the deployment map"},{"kind":"claim","claim":"agrep -l lists matching chats, not every line","source":"# matching chats, not every line","input":"lantern","expected":"one chat line for the fixture session","probe_kind":"cli","argv":["python3","cli.py","-l","lantern"],"cwd":"<guest-path>","env":{"AGREP_HOME":"/tmp/agrep-home"},"expected_text":"upgrade-fixture"},{"kind":"claim","claim":"agrep around shows context around a hit","source":"# context around a hit","input":"session 11111111-1111-4111-8111-111111111111 turn 1","expected":"turn context including turn 1 text","probe_kind":"cli","argv":["python3","cli.py","around","11111111-1111-4111-8111-111111111111","1"],"cwd":"<guest-path>","env":{"AGREP_HOME":"/tmp/agrep-home"},"expected_text":"Both synthetic checklists are stored together."}]
EOF
cat > examples.json <<'EOF'
[{"kind":"example","claim":"Development quickstart keyword search runs from the source checkout","source":"README.md Development section","input":"copper lantern","expected":"fixture hit printed","argv":["python3","cli.py","copper lantern"],"cwd":"<guest-path>","env":{"AGREP_HOME":"/tmp/agrep-home"},"expected_text":"copper lantern upgrade beacon beside the deployment map"},{"kind":"example","claim":"Filter by agent with --agent","source":"README.md The commands you actually need section","input":"copper lantern --agent claude","expected":"hits from the claude agent","argv":["python3","cli.py","copper lantern","--agent","claude"],"cwd":"<guest-path>","env":{"AGREP_HOME":"/tmp/agrep-home"},"expected_text":"copper lantern upgrade beacon beside the deployment map"}]
EOF
python3 -m json.tool claims.json >/dev/null && python3 -m json.tool examples.json >/dev/null && echo ok; cd <guest-path> && <env-var>=<guest-path> python3 -I <guest-path> --validate --replay > /tmp/validate2.log 2>&1; echo EXIT=$?; tail -c 1800 /tmp/validate2.log
cd <guest-path> && grep -o '"verdict": "[a-z]*"' /tmp/validate2.log | sort | uniq -c; grep -o '"valid": [a-z]*' /tmp/validate2.log | head -2; echo ---; git init -q . && git remote add origin https://github.com/dannyisbad/agrep && git fetch --depth 1 origin ac4efadbc0a305ac8bbb068422c5437157f35f56 2>&1 | tail -3; git reset --mixed FETCH_HEAD 2>&1 | tail -2; git rev-parse HEAD
cd <guest-path> && cat > report.json <<'EOF'
{"outcome":"working","fault_owner":null,"operation":"Keyword search over indexed agent transcripts from the source checkout: python3 cli.py \"copper lantern\" returns the matching fixture messages, plus -E regex, -l chat listing, around context and --agent filtering.","operation_scope":"Linux (WSL2 kernel 6.6), Python 3.12.3 (README requires 3.10+), Rust 1.97.0 via rustup per rust-toolchain.toml, Node 22 present but unused. Searched a copy of the repo's bench/fixtures/upgrade_store Claude transcripts placed at AGREP_HOME=/tmp/agrep-home, because the guest has no real agent history. No semantic model, no agent hooks, no instruction teaching.","steps":["git init; git remote add origin https://github.com/dannyisbad/agrep; git fetch --depth 1 origin ac4efadbc0a305ac8bbb068422c5437157f35f56; git reset --mixed FETCH_HEAD","curl -sSf https://sh.rustup.rs -o /tmp/rustup-init.sh && sh /tmp/rustup-init.sh -y --profile minimal --default-toolchain 1.97.0 -c clippy -c rustfmt (installs toolchain from rust-toolchain.toml)","python3 cli.py setup --yes --no-teach --no-hook --no-semantic (README Development: python cli.py setup; ingest binary built with cargo build --release, 2m12s)","AGREP_HOME=/tmp/agrep-home python3 cli.py \"copper lantern\" (README: agrep \"search term\" equivalent)","AGREP_HOME=/tmp/agrep-home python3 cli.py status --json (index_state ready, messages 2)","AGREP_HOME=/tmp/agrep-home python3 cli.py -l lantern; python3 cli.py around <id> 1; python3 cli.py zzqxvnotpresentterm (exit 1)"],"deviations":["System cargo was 1.75 while rust-toolchain.toml pins 1.97.0; installed rustup 1.97.0 into /var<guest-path> The first rustup run printed unrelated cc/cargo build lines but a second run confirmed the install.","setup was run with --yes but without --no-archive, so setup enabled archive retention in <guest-path> This was an unintended default; disposable guest only.","First keyword search right after setup returned hits but exited 2 with 'first-run completion could not verify the published snapshot' while the index built in the background. A rerun after ~20s exited 0 with index_state ready.","The guest had no agent transcripts, so the repo's bench/fixtures/upgrade_store/claude fixture was copied to /tmp/agrep-home/.claude/projects and read through AGREP_HOME.","The README's exit-1 statement for an exhaustive keyword miss was observed manually (zzqxvnotpresentterm exited 1 with '0 matching rows') but was not kept as a claim: the checker only validates nonzero exit codes against an exit-code table, and the repo docs have none. Replaced it with a zero-exit regex claim.","Not run: bench/perf.py --check and bench/semantic_cli_scale.py --check (timing budgets, not needed for functional proof), the Rust/Python test suites, and the semantic/ONNX lane.","The product's freshness daemon (pid reported by status --json) remains running from the product's own on-demand refresh; it was not started by the checker."],"stuck_step":null,"main_blocker":null,"plain_failure":null,"suggested_fix":"none"}
EOF
python3 -m json.tool report.json >/dev/null && echo report-ok; cd <guest-path> && <env-var>=<guest-path> python3 -I <guest-path> --validate --replay > /tmp/validate3.log 2>&1; echo FINAL_EXIT=$?; grep -o '"valid": [a-z]*' /tmp/validate3.log | head -1; grep -o '"evidence_strength": "[a-z_]*"' /tmp/validate3.log