Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion .claude/skills/deep-code-review/INDEX.md
Original file line number Diff line number Diff line change
Expand Up @@ -138,7 +138,7 @@ Generated by `scripts/skill_index.py` (do not hand-edit; rerun it after changing
| `references/testing-ui.md` | the target renders a UI or ships browser / E2E specs. | 4868 |
| `references/threat-modeling.md` | a change adds a trust boundary, an auth path, a data flow to a new party, an agent or tool capabili… | 2574 |
| `references/time-date-correctness.md` | reviewing anything that stores, computes, schedules, or renders timestamps, durations, recurring ev… | 4809 |
| `references/tool-hub.md` | a review should lean on external scanners via the tool hub (workflows, Python, Go, SBOM/lockfiles,… | 11335 |
| `references/tool-hub.md` | a review should lean on external scanners via the tool hub (workflows, Python, Go, SBOM/lockfiles,… | 12426 |
| `references/ux-components.md` | the target or diff renders one concept in more than one place: a shared component or design token,… | 7235 |
| `references/ux-dataviz.md` | the target or diff renders a chart, sparkline, heat-map or calendar grid, a KPI / metric delta, a c… | 2787 |
| `references/ux-gates.md` | a UI change is claimed verified on screenshot evidence, the project's auto-merge or CI policy gates… | 3756 |
Expand Down
5 changes: 5 additions & 0 deletions .claude/skills/deep-code-review/references/tool-hub.md
Original file line number Diff line number Diff line change
Expand Up @@ -60,7 +60,12 @@ The agent sandbox denies installs and most network, so the owner provisions scan
| `detect-secrets` | Potential secrets in any changed file, no baseline: one scan of the changed files, one row per hit (rule = the plugin type, message = `potential secret (value not printed)`). The `hashed_secret` and any value are never copied into a row, the log or the report. Skips non-JSON log lines around the report. Empty or whitespace output, a crash, any nonzero exit (`scan` exits 0 whatever it finds, so a nonzero exit is a failure) or a report without a `results` object is `not run`, never clean. A heuristic detector beside `gitleaks`, so expect an overlap and some false positives | `detect-secrets scan --no-verify FILE...` (`--no-verify` turns off the additional network verification; no baseline is read or written; the JSON report goes to stdout) | verified (README: `scan [path ...]`, `--no-verify`, `--all-files`; `main.py`: the scan action never returns nonzero; `secrets_collection.py` and `potential_secret.py`: results keyed by filename, each hit with `type`, `filename`, `hashed_secret`, `is_verified`, `line_number`; `baseline.py`: `results` is a top-level key), fetched 2026-11-12 | route-only |
| `njsscan` | Node.js source: a changed `.js`, `.jsx`, `.mjs` or `.cjs` file. ERROR and WARNING rules in both the `nodejs` and `templates` sections become rows (rule = rule id, message = severity and description, line = first `match_lines` entry), INFO rules are dropped. Skips non-JSON log lines around the report. A report without `nodejs`, `templates` and `errors`, a non-empty `errors` list, empty or whitespace output, a crash, exit 0 with an ERROR row, or exit 1 with no ERROR row is `not run`, never clean. The README says nothing about network use; `__main__.py` imports no network module, which is not a claim about the scan modules | `njsscan --json FILE...` (JSON to stdout when `-o` is omitted; `-w` is not used, so WARNING rules do not change the exit code). Exit 1 = an ERROR rule matched or files were not scanned | verified (README: `--json`, `-o`, `-w`, `path ...`, result keys `nodejs`, `templates`, `errors` with `files` and `metadata`; `__main__.py` `handle_exit`: exit 1 on errors or an ERROR rule; `json_out.py`: `print` to stdout without `-o`), fetched 2026-11-12 | route-only |
| `sobelow` | Elixir Phoenix apps: a changed `.ex`, `.exs`, `.eex`, `.heex` or `.leex` file, only when the root `mix.exs` names `{:phoenix,` and neither `mix.exs` nor `.sobelow-conf` is changed in the diff (a project config is code: that run is NOT CHECKED). Scans the whole app and keeps rows on changed files; the confidence tier (high, medium, low) is the message, rule = the finding type. Skips non-JSON log lines around the report. Empty or whitespace output, a crash, a report whose lists or `total_findings` are missing or disagree, exit 0 with a finding, or exit 1 with none is `not run`, never clean. A non-Phoenix project prints text and exits 0 with no JSON, so it is `not run`. `--private` skips the update check (the tool otherwise fetches a version file from its own site); no other network use was found in `lib/sobelow.ex` | `sobelow --format json --private --exit low` (JSON to stdout; `--exit low` returns 1 when any finding exists) | verified (README: `--format`, `--exit`, `--private`, "All `json` formatted findings contain a `type`, `file`, and `line` key", escript install; `lib/sobelow.ex`: `exit_with_status` halts 1 when high, medium or low findings exist, the non-Phoenix path prints text and halts 0, the update check is skipped when `private`; `lib/sobelow/finding_log.ex`: `json` emits `findings` with `high_confidence`, `medium_confidence`, `low_confidence`, plus `total_findings`; `mix.exs`: Apache-2.0), fetched 2026-11-12; unverified: the findings' other keys | route-only |
| `flawfinder` | C/C++ source (`.c .h .cc .cpp .cxx .hpp .hh .hxx`), changed files in one call; rows from the SARIF results (rule, file, line; the snippet is never read); wired only once installed | `flawfinder --sarif --quiet FILES`; `main()` returns 0, or 1 only with `--error-level` (not passed), or 15 on encoding errors, so any non-zero exit, empty output or a document without `runs[].results` is not run; a clean scan prints the document with `"results": []` | verified (man page: `--sarif`, `--quiet`, `--error-level`; source at c4216b74cf: `output_sarif`, `flawfind()` and the exit values; GPLv2+, the OSI list has GPL v2). Egress: the source imports no network module (only `urllib.parse.quote`); the SARIF holds schema and taxonomy URLs as text, never fetched | route-only |
| `cppcheck` | C/C++ source, changed files in one call, no project build and no include paths, so default checks only (no `--enable`: its values were not verified here); `information` notes (missing includes) are dropped, any other finding needs a file and line; wired only once installed | `cppcheck --xml-version=2 --quiet FILES`; the XML goes to stderr (`cppcheck.1.xml`: "Write results in XML to error stream"), which the runner merges into its capture; no `--error-exitcode`, so the exit is 0 (`settings.exitCode`) and any other exit, missing `<results><errors>` or crash is not run. The primary location is the first `<location>` (manual: "The primary location is listed first"; `toXML` writes the call stack reversed), later ones are context | verified (manual and man page; source at 103143dcb6: `--xml-version` accepts 2 or 3, `ErrorMessage::toXML`, `--error-exitcode`; COPYING is GPLv3, on the OSI list). Egress: no curl, socket or HTTP client call in `cli/`, `lib/` or `frontend/` (grep of the cloned source); the `--xml` sample is the manual's | route-only |
| `devskim` | Multi-language source (Microsoft DevSkim, MIT), one changed file per run (`analyze --source-code` takes a file: `AnalyzeCommand.cs` has a single-file branch); rows from the SARIF results; wired only once installed | `devskim analyze --source-code FILE --file-format sarif`; SARIF to stdout (`-O` unset, `Console.Out`); without `-E` the exit is 0 whatever it finds, and `ExitCode.cs` makes errors negative (`CriticalError` -2, `ArgumentParsingError` -3), so any non-zero exit, empty output or a document without `runs[].results` is not run. `--file-format` is passed because the writer factory falls back to `text` when it is empty | verified for flags, exit codes and license (wiki Analyze Command; source at 7f8e5e24f6: options in `BaseAnalyzeCommandOptions.cs`, exit codes, `SarifWriter.cs`; README: MIT), not for the SARIF shape. Egress: no HTTP client, update or telemetry code in `DevSkim-DotNet` `*.cs` (grep); its NuGet dependencies (CommandLineParser, LibGit2Sharp, Sarif.Sdk, ApplicationInspector, Newtonsoft.Json) were not audited, so egress by a dependency is `unverified`; the SARIF help URIs are text. The SARIF shape (`runs[].results[]`) is stub-tested only, not verified: no real DevSkim output was fetched or run, so a real report may differ | route-only |
| `guarddog` | npm packages a changed `package-lock.json` newly adds (name@version absent from the merge-base lock; a bump counts) and that are installed under `node_modules` at the locked version (the installed `package.json` says so); `hasInstallScript` ones first, then at most 25 and about 300 s in all (`REVIEW_GUARDDOG_BUDGET`), the rest named in a `not-checked` row; not under `--full`; not a `hub=` row | `guarddog npm scan --output-format json DIR` on existing in-repo dirs only (it fetches a non-directory argument as a registry name), only when `guarddog --version` says 3.x; `--no-sandbox` (behind a dead proxy) only after its sandbox fails to start (nested); never `verify`; leads: `threat-*` rules only, each tagged `[heuristic]`; a rule that errors is a `not-checked` row and the other rules' hits stay; a newly added lock entry with `hasInstallScript` is a `package-lock` lead of its own | verified (`--help`, 3.2.0 source, real run 2026-10-30); the install-hook rules (`threat-npm-preinstall-script`, `threat-process-hooks`) match `*/package.json`, a nested one, so a package's own root `package.json` hook is never flagged | route-only |
| `detekt` | Not routed: Kotlin static analysis, Apache-2.0 (https://github.com/detekt/detekt LICENSE, on the OSI list at https://opensource.org/license/apache-2-0, fetched 2026-11-12). The CLI docs (https://detekt.dev/docs/gettingstarted/cli, fetched 2026-11-12) give `--input`, `--report sarif:PATH` (the XML-like format is `checkstyle`) and exits 0 clean, 1 unexpected error, 2 issues found, 3 invalid config. `Main.kt` and `CliArgs.kt` hold no HTTP, update or telemetry code, but the analysis core, the Kotlin compiler and plugin loading were not read, so its network-free behavior is `unverified` (not checked = not routed); it is also a code-quality linter whose default rule sets are not security rules | not run | not routed | not routed |
| `security-code-scan` | Not routed: a Roslyn analyzer (NuGet package or Visual Studio extension, LGPL-3.0 per the repository page, https://github.com/security-code-scan/security-code-scan, fetched 2026-11-12) that runs inside a project build; the stand-alone runner was not verified here and a build needs a NuGet restore (network). It cannot scan changed files only, so it is not routed; `devskim` is the routed .NET source scanner | not run | not routed | not routed |
| `bearer` | Not routed: Bearer CLI is under the Elastic License 2.0, which forbids offering the software to third parties as a hosted or managed service (https://github.com/Bearer/bearer, fetched 2026-11-12), and that license is not on the OSI approved-licenses list (https://opensource.org/licenses, fetched 2026-11-12), so it is not routed | not run | not routed | not routed |
| `snyk-agent-scan` | MCP and skill configs | not run | unverified: docs show no offline mode (it calls the Agent Scan API and can start MCP servers) | route-only |
| `opengrep` | SAST fallback: stands in for `semgrep` when `semgrep` is absent and `opengrep` is installed; same rules dir (repo config, else the bundled rules), never the registry pack; rows keep tool `semgrep` and the closing line says `opengrep` ran | `opengrep scan --config DIR --json --quiet -- FILES` (env `SEMGREP_SEND_METRICS=off`) | unverified (its README shows `scan` and `-f` only; `--config` and `--json` are inherited from semgrep, not confirmed); a crash is `not run`, never clean | route-only, local |
Expand Down
Loading
Loading