Skills/PyPI/symbols-mcp

symbols-mcp@3.14.786

CAUTION

Publisher claims: "Symbols.app MCP — docs, code generation, publishing, CLI/SDK reference" Observed in sandbox: no network, touches nothing outside the project, writes no memory.

Registry PyPI (MCP server)Publisher pypi · publishing for 728 daysInstalls 3,125Last detonated 18 h ago

Coverage

Server started and answered tools/list.
Tested
MCP initialize and tools/list · every tool description, for hidden instructions · every file read, file written and host contacted by what ran
Not tested
what the tools do when called with real credentials

Before you give it your token

It will ask for
SYMBOLS_TOKENConfiguration, not credentials: FRANK_AUDIT_BIN, FRANK_AUDIT_URL, MCP_HOST, MCP_PORT, MCP_TRANSPORT, SYMBOLS_API_URL, SYMBOLS_SKILLS_DIR
Where it talks to
symbols.app (its own)Also esm.sh, symbo.ls — not the publisher's domain; check what for before connecting.
What we saw when it ran
no credential files read · no network
What we could not test
what the tools do when called with real credentials

Connect it with a scoped credential, under the guard, and read the third-party hosts first. The code names esm.sh, symbo.ls as well as its own vendor. The guard injects SYMBOLS_TOKEN only into requests to symbols.app (and its subdomains) and refuses the rest.

$vet-skills run --secrets SYMBOLS_TOKEN=symbols.app -- uvx symbols-mcp

Never paste a credential into a chat, a ticket or a report, including ours. How the guard holds it.

What changed since 3.14.784

vet-skills diff symbols-mcp 3.14.784 3.14.786

This version reaches for something 3.14.784 did not: 9 rewritten tool descriptors (audit_component, convert_html, convert_react, detect_environment, …). Compared from the two recorded runs, not from source.

Rewritten tool descriptors
audit_component, convert_html, convert_react, detect_environment, generate_component, generate_page, get_cli_reference, get_project_rules, get_sdk_reference

The guard can hold an update like this until someone approves it (runtime.on_change: hold). A change that only shows with a real credential, or on a path the run did not reach, is not in a trace and so not in this comparison.

Is [email protected] safe to install?

KillerSkills ran it in the KillerSkills honeypot sandbox (2 scenarios) on 2026-11-19 and scored it 61/100 (Caution). It is not known to be malicious, but it has behaviour or code a security policy should decide on before install. Notable findings: Instructs the agent to modify its persistent memory or instruction files; Access to sensitive file paths; URL with embedded credentials. Summary: no network, touches nothing outside the project, writes no memory.

Behavior trace

Honeypot run · 20 s · 2 scenarios · 0 deviations
t+CategoryObserved action
0.0sprocessno install scripts in bundle
20sclaimed18 tools: get_project_rules, search_symbols_docs, get_cli_reference, get_sdk_reference, generate_component, generate_page, convert_react, convert_html, detect_environment, audit_component, convert_to_json, login, list_projects, create_project, get_project, save_to_project, publish, push
20stoolsf201980daa1ff6c9
20stoolsb39e809295ca146c
20stoolsf3913938f66f80f7
20stoolsa87e22be7ab80ebe
20stools18348e9ddbf6cacc
20stools38a298164f55fa2d
20stoolsdbde373d0e61be79
20stools6710358245019e35
20stools781f38a39bb13fd2
20stools87c1821f7b52b381
20stools0d5d6d6e553eb1bb
20stools2edce777e1ad8480
20stools3b9e2202d0be6779
20stoolsb15c38a8ddb171f8
20stools84efd9ae6100b537
20stoolsd91a149bd20d6b17
20stoolsbcc4e7cd97211473
20stoolsfe60d8aa3bee6068
20sprocessscript backend runs setup only
20sprocessno install scripts to re-run

Static findings

HIGH
Instructs the agent to modify its persistent memory or instruction filesAGENTS.md:1 · Instructs the agent to modify its persistent memory or instruction files<!-- generated from CLAUDE.md (canonical) 2026-08-20 — edit CLAUDE.md, then re-copy -->Fix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
HIGH
Access to sensitive file pathssymbols_mcp/skills/FRANKABILITY.md:543 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files… pre-boot code, orphans, `files/` manifests and comments; not dotfiles such as `.env`, where a local secret belongs: - known key formats — AWS access key ids (`AKI…Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
HIGH
URL with embedded credentialssymbols_mcp/skills/DESIGN_SYSTEM.md:1242 · URL with embedded credentials — may be phishing or credential leakingurl: 'https://fonts.googleapis.com/css2?family=Inter:[email protected]&display=swap',Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST07
CRITICAL
Rule modification attemptsymbols_mcp/skills/MIGRATION.md:251 · Rule modification attempt — overwriting existing instructions### Rule 4: Replace `$propsCollection` with `children`Fix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
HIGH
Auto-update or dynamic fetch configurationsymbols_mcp/skills/COOKBOOK.md:223 · Auto-update or dynamic fetch configuration — agent may load new code at runtimeAuto-update every second using setInterval.Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
HIGH
Authorization escalation languagesymbols_mcp/skills/LEARNINGS.md:765 · Authorization escalation languageis the ITEM STATE, so `item.root` gives full root access. `this` (bound byFix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
CRITICAL
Prompt injectionsymbols_mcp/skills/DESIGN.md:205 · Prompt injection — explicit attempt to bypass safety measures- Avoid content collapse that hides key functionsFix: Keep SKILL.md to visible, descriptive text. Remove comments and encoded text that address the agent.
AST01
HIGH
document.cookie accesssymbols_mcp/skills/RULES.md:2116 · document.cookie access — can steal session cookiesdocument.cookie = '…' // wrap in a `functions/` helper, never inlineFix: Replace encoded or assembled strings with plain literals and remove undeclared network calls.
AST07
CRITICAL
Rule modification attemptsymbols_mcp/skills/MIGRATION.md:180 · Rule modification attempt — overwriting existing instructions### Rule 2: Rename `extend` to `extends`, `childExtend` to `childExtends`Fix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
CRITICAL
Rule modification attemptsymbols_mcp/skills/MIGRATION.md:153 · Rule modification attempt — overwriting existing instructions### Rule 1: Flatten `props` and `on`Fix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
CRITICAL
Rule modification attemptsymbols_mcp/skills/MIGRATION.md:276 · Rule modification attempt — overwriting existing instructions### Rule 5: Replace `props: (fn)` function with individual prop functionsFix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
HIGH
Sends the whole process environment outsymbols_mcp/server.py:2806 · Sends the whole process environment out"local": "Local Symbols project. Run with `smbls start` (parcel/vite/esbuild/webpack — see bundler field). Push with `smbls push <env>`.",Fix: Never read or enumerate credential locations; accept a documented environment variable instead.
AST03
CRITICAL
Rule modification attemptsymbols_mcp/skills/MIGRATION.md:192 · Rule modification attempt — overwriting existing instructions### Rule 3: Replace `$stateCollection` with `children` + `childrenAs: 'state'`Fix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
MEDIUM
Python subprocess execution detectedtests/test_post_write_hook.py:133 · Python subprocess execution detectedproc = subprocess.run(["bash", str(HOOK)], input=payload, capture_output=True,Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
MEDIUM
Python subprocess execution detectedtests/test_search_docs.py:115 · Python subprocess execution detectedproc = subprocess.run(["node", str(SERVER)], input="\n".join(lines) + "\n", capture_output=True, text=True, timeout=120)Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
MEDIUM
npx invocation of non-well-known packagebin/symbols-mcp.cjs:10 · npx invocation of non-well-known package — may execute untrusted code// Subcommand dispatch — `npx @symbo.ls/mcp <subcommand>` shortcuts.Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
Python subprocess execution detectedtests/test_node_parity.py:37 · Python subprocess execution detectedproc = subprocess.run(["node", str(NODE_SERVER)], input="\n".join(lines) + "\n",Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
MEDIUM
File type combined with upload actionsymbols_mcp/skills/CLI.md:50 · File type combined with upload action — agent may exfiltrate local data files…e id is a 12-character hash of the symbols `dir`, the `entry` document, `symbols.json`, `package.json`, the nearest lockfile and the CLI version. An id that does not match `/^[A-Za-z0-9._-]{1,22}$/` stops the build before Parcel starts. Th…Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
npx invocation of non-well-known packagesymbols_mcp/server.py:262 · npx invocation of non-well-known package — may execute untrusted code"`.claude/hooks/`, suggest the user run `npx @symbo.ls/mcp init-rules` (or equivalent) to drop "Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
File type combined with upload actionsymbols_mcp/skills/FRAMEWORK.md:1016 · File type combined with upload action — agent may exfiltrate local data files… `brender.paths` lists addresses for them. Both run one route loop: the `symbols.json` `brender` settings (`context`, `exclude`, `paths`, `strict`, `eagerImages`, `timeout`, `stamp`), `file:` URLs mapped into the output, and the same diagn…Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
Python subprocess execution detectedtests/test_post_write_hook.py:56 · Python subprocess execution detectedproc = subprocess.run(Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
MEDIUM
File type combined with upload actionsymbols_mcp/skills/CLI.md:174 · File type combined with upload action — agent may exfiltrate local data files…e project loads are resolved; `"frank": { "importMetaUrls": "all" }` in `symbols.json` also resolves calls inside function bodies and in `app.js`. `smbls frank to-json` inlines such files as `data:` URLs and never uploads; `sync`, `collab`…Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
File type combined with upload actionbin/symbols-mcp-init-rules.cjs:340 · File type combined with upload action — agent may exfiltrate local data filesconsole.log(` • PostToolUse runs frank-audit + FA-rule check after every JS edit`)Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
File type combined with upload actionsymbols_mcp/skills/CLI.md:178 · File type combined with upload action — agent may exfiltrate local data files…out a verified project id (run `smbls push` or `smbls sync` first to stamp `lock.json`). Its websocket handshake then names the project by `projectId` **and** `owner` (from `symbols.json`'s `owner` field) as a backstop — never a bare key, …Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
Python subprocess execution detectedsymbols_mcp/server.py:865 · Python subprocess execution detectedproc = subprocess.run(Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
MEDIUM
npx invocation of non-well-known packagesymbols_mcp/skills/CLI.md:110 · npx invocation of non-well-known package — may execute untrusted codenpx -y @symbo.ls/mcp init-rules --force # overwrite existing filesFix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packagesymbols_mcp/skills/CLI.md:109 · npx invocation of non-well-known package — may execute untrusted codenpx -y @symbo.ls/mcp init-rules # idempotent; project-levelFix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
Python subprocess execution detectedtests/test_node_audit_component.py:45 · Python subprocess execution detectedproc = subprocess.run(Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
MEDIUM
File type combined with upload actionbin/symbols-mcp-init-rules.cjs:341 · File type combined with upload action — agent may exfiltrate local data filesconsole.log(` Disable individually: SYMBOLS_MCP_REQUIRE_RULES=0 / SYMBOLS_MCP_REMINDER=0 / SYMBOLS_MCP_POST_AUDIT=0`)Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
Python subprocess execution detectedtests/test_node_cli_reference.py:29 · Python subprocess execution detectedproc = subprocess.run(Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
MEDIUM
npx invocation of non-well-known packagesymbols_mcp/skills/CLI.md:97 · npx invocation of non-well-known package — may execute untrusted code…r to explicit deps ``` Both `smbls init` and `smbls create` automatically run `npx -y @symbo.ls/mcp init-rules` at the end of scaffolding — drops `CLAUDE.md`, `AGENT…Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packagesymbols_mcp/skills/CLI.md:111 · npx invocation of non-well-known package — may execute untrusted codenpx -y @symbo.ls/mcp init-rules --global # user-global ~/.claude/ insteadFix: Pin versions and sources; vendor anything fetched at install time.
AST06
HIGH
Authorization escalation languagesymbols_mcp/skills/SDK.md:1040 · Authorization escalation language### Admin overridesFix: Remove the text that addresses the agent rather than the reader; describe behavior, do not instruct.
AST01
MEDIUM
File type combined with upload actionbin/symbols-mcp.cjs:534 · File type combined with upload action — agent may exfiltrate local data files- Reverse with: \`smbls frank to-fs data.json -o ./output\` or \`toFS(data, './output')\`Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
npx invocation of non-well-known packagesymbols_mcp/skills/AUDIT.md:739 · npx invocation of non-well-known package — may execute untrusted codenpx -y -p @symbo.ls/mcp symbols-audit --json ./symbolsFix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
Python subprocess execution detectedtests/test_audit_component.py:244 · Python subprocess execution detectedr = subprocess.run([server.FRANK_AUDIT_BIN, "help"], capture_output=True, text=True, timeout=60)Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
HIGH
document.cookie accesssymbols_mcp/skills/RULES.md:2195 · document.cookie access — can steal session cookies… `globalThis.`, `document.title`, `document.body`, `document.documentElement`, `document.cookie`, `document.createElement`, `document.querySelector`, `document.getElementById`…Fix: Replace encoded or assembled strings with plain literals and remove undeclared network calls.
AST07
MEDIUM
File type combined with upload actionsymbols_mcp/skills/CLI.md:176 · File type combined with upload action — agent may exfiltrate local data files…ll')` in a handler). A package nothing declares still fails to resolve (`package.json` dependencies never reach the platform's importmap), and a value computed from such a stub while the project loads fails the build (`FRANK_UNINSTALLED_…Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
Process kill commandsymbols_mcp/skills/FRAMEWORK.md:1534 · Process kill command — agent can terminate processesThen restart any active dev server. Don't kill the portless proxy onFix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
npx invocation of non-well-known packagesymbols_mcp/skills/AUDIT.md:735 · npx invocation of non-well-known package — may execute untrusted codenpx -y -p @symbo.ls/mcp symbols-audit ./symbolsFix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
Python subprocess execution detectedtests/test_node_stdio_exit.py:28 · Python subprocess execution detectedproc = subprocess.run(["node", str(SERVER)], input="\n".join(lines) + "\n",Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
MEDIUM
Python subprocess execution detectedtests/test_runtime_examples.py:74 · Python subprocess execution detectedproc = subprocess.run(["node", str(RUNNER), TREE], capture_output=True, text=True, timeout=300)Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
MEDIUM
Python subprocess execution detectedtests/test_structural_checks.py:390 · Python subprocess execution detectedproc = subprocess.run(["node", str(NODE_SERVER)], input="\n".join(lines) + "\n",Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
MEDIUM
npx invocation of non-well-known packagesymbols_mcp/skills/CLI.md:15 · npx invocation of non-well-known package — may execute untrusted codenpx @symbo.ls/cli <command>Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
Python subprocess execution detectedsymbols_mcp/server.py:46 · Python subprocess execution detectedresult = subprocess.run(Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
MEDIUM
File permission modification commandpublish.sh:63 · File permission modification command — agent changes file ownership/permissionschmod +x bin/symbols-mcp.jsFix: Request the narrowest permission that works; replace wildcards with the exact paths or hosts.
AST05
MEDIUM
Python subprocess execution detectedtests/test_frame_calls.py:244 · Python subprocess execution detectedproc = subprocess.run(["node", str(NODE_SERVER)], input="\n".join([json.dumps(init), *lines]) + "\n",Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
MEDIUM
Python subprocess execution detectedtests/test_ambiguous_project_key.py:105 · Python subprocess execution detectedproc = subprocess.run(["node", str(NODE_SERVER)], input="\n".join(lines) + "\n",Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
MEDIUM
Python subprocess execution detectedtests/test_npm_bins.py:37 · Python subprocess execution detectedr = subprocess.run(Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
MEDIUM
npx invocation of non-well-known packagesymbols_mcp/skills/AUDIT.md:8 · npx invocation of non-well-known package — may execute untrusted code> - From a shell (CI / pre-commit): `npx -y -p @symbo.ls/mcp symbols-audit ./symbols`Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
File type combined with upload actionsymbols_mcp/server.py:2689 · File type combined with upload action — agent may exfiltrate local data filesoutput += "- Reverse with: `smbls frank to-fs data.json -o ./output` or `toFS(data, './output')`\n"Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
Python subprocess execution detectedtests/test_runtime_examples.py:59 · Python subprocess execution detectedproc = subprocess.run(["node", str(NODE_SERVER)], input="\n".join(lines) + "\n",Fix: Build commands from fixed arguments, never from interpolated strings; prefer an API client over shelling out.
AST05
MEDIUM
npx invocation of non-well-known packagesymbols_mcp/skills/AUDIT.md:737 · npx invocation of non-well-known package — may execute untrusted codenpx -y -p @symbo.ls/mcp symbols-audit --strict ./symbolsFix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
File type combined with upload actionsymbols_mcp/skills/CLI.md:350 · File type combined with upload action — agent may exfiltrate local data files…--strict`, `--out-dir <dir>`, `--bundler`. `build` copies `public/` (or `symbols.json` `publicDir`) into the output root and never replaces a file the bundler wrote. With brender on, `build` pre-…Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
File type combined with upload actionsymbols_mcp/skills/AUDIT.md:122 · File type combined with upload action — agent may exfiltrate local data files├── after.snapshot.json # post-audit metrics (final iteration)Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
npx invocation of non-well-known packagetemplates/skills/symbols/SKILL.md:52 · npx invocation of non-well-known package — may execute untrusted code- `npx -y @symbo.ls/mcp init-rules` — (re)install these rules + skills for every detected agentFix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packagetemplates/agent-rules/CLAUDE.md:205 · npx invocation of non-well-known package — may execute untrusted codenpx -y -p @symbo.ls/mcp symbols-audit ./symbolsFix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
File type combined with upload actionsymbols_mcp/skills/FRAMEWORK.md:1190 · File type combined with upload action — agent may exfiltrate local data files… installed package still bundles inline; an undeclared one still fails (`package.json` dependencies do not count); a value computed from the stub while the project loads fails with `FRANK_UNINSTALLED_DEPENDENC…Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
npx invocation of non-well-known packagetests/test_npm_bins.py:3 · npx invocation of non-well-known package — may execute untrusted codenpx runs the bin whose name is the package name without its scope (`mcp`);Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packagetests/test_npm_bins.py:1 · npx invocation of non-well-known package — may execute untrusted code"""`npx @symbo.ls/mcp <subcommand>` needs a bin named after the package.Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
Process kill commandsymbols_mcp/skills/RULES.md:1612 · Process kill command — agent can terminate processes# restart only the app's direct Parcel port (NEVER kill :1355 — that's the shared portless proxy)Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
npx invocation of non-well-known packagetemplates/agent-rules/CLAUDE.md:221 · npx invocation of non-well-known package — may execute untrusted code> If `symbols-mcp` is not registered in your MCP config, see the project's SETUP.md (or `npx @symbo.ls/mcp` to start the server). Without the MCP, every code suggestion is guesswork.Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packagetemplates/agent-rules/CLAUDE.md:17 · npx invocation of non-well-known package — may execute untrusted code…`mcp__symbols-mcp__audit_project()` to get the multi-phase playbook, pair with `npx -y -p @symbo.ls/mcp symbols-audit ./symbols` (the CLI), iterate until convergence …Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packagesymbols_mcp/skills/RUNNING_APPS.md:446 · npx invocation of non-well-known package — may execute untrusted code2. **Cloudflare Worker**: `npx wrangler deploy` (edge deployment)Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
Access to sensitive file pathstests/test_node_audit_component.py:87 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files("../../etc/passwd.js", "components/cardProps.js"),Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
MEDIUM
npx invocation of non-well-known packagetests/test_npm_bins.py:6 · npx invocation of non-well-known package — may execute untrusted code./symbols` and `npx -y @symbo.ls/mcp init-rules`, as documented, failed.Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
File type combined with upload actionsymbols_mcp/skills/RUNNING_APPS.md:364 · File type combined with upload action — agent may exfiltrate local data filessmbls frank to-fs project.json -o ./outputFix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
File type combined with upload actionsymbols_mcp/skills/FRAMEWORK.md:1200 · File type combined with upload action — agent may exfiltrate local data files… When editing any project whose published artifact is a frank-bundled JSON (`app.json` or the referenced data inside `symbols.json`), the compiled JSON becomes stale on every source edit. **Always run `frank` to recompile** after any edit …Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
npx invocation of non-well-known packagetemplates/agent-rules/AGENTS.md:11 · npx invocation of non-well-known package — may execute untrusted code5. Full audit: `audit_project()` playbook + `npx -y -p @symbo.ls/mcp symbols-audit ./symbols` CLI.Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
Access to sensitive file pathstests/test_audit_component.py:225 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system filesout = server.audit_component(COMPONENT_FILE, file_path="../../etc/passwd.js")Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
MEDIUM
File type combined with upload actionsymbols_mcp/skills/WORKSPACE.md:52 · File type combined with upload action — agent may exfiltrate local data files…m '../<app>/symbols/context.js'` - `smbls libs link` reads the target's `symbols.json` to compute the right import path. Pre-3.14.35 CLI ignored `dir` and hardcoded `<targetRoo…Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
Process kill commandsymbols_mcp/skills/SDK.md:938 · Process kill command — agent can terminate processesUser-facing reads + admin CRUD. Full recipes (kill switch, allowlist,Fix: Build commands from fixed arguments; do not interpolate untrusted strings into a shell.
AST05
MEDIUM
npx invocation of non-well-known packagetests/test_npm_bins.py:8 · npx invocation of non-well-known package — may execute untrusted codenamed bins stay for `npx -p @symbo.ls/mcp <bin>` and global installs.Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
File type combined with upload actionsymbols_mcp/skills/FRAMEWORK.md:1075 · File type combined with upload action — agent may exfiltrate local data files…rocessed config and the global CSS). The build writes it once as `smbls-ds.<sha>.json` next to the bundles and puts `<link rel="preload" as="fetch" data-smbls="design-system">` into every page it pre-renders (their boot waits for it). The …Fix: Remove the call, or declare the host in frontmatter and explain in the README what is sent and why.
AST07
MEDIUM
Access to sensitive file pathstests/test_audit_component.py:106 · Access to sensitive file paths — SSH keys, AWS credentials, env files, or system files@pytest.mark.parametrize("bad", ["../../etc/passwd.js", "functions/../../x.js", ".."])Fix: Stop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
AST05
MEDIUM
npx invocation of non-well-known packagetests/test_post_write_hook.py:32 · npx invocation of non-well-known package — may execute untrusted codenpx = bin_dir / "npx"Fix: Pin versions and sources; vendor anything fetched at install time.
AST06
MEDIUM
npx invocation of non-well-known packagetests/test_npm_bins.py:5 · npx invocation of non-well-known package — may execute untrusted codedetermine executable to run" — so `npx -y @symbo.ls/mcp symbols-auditFix: Pin versions and sources; vendor anything fetched at install time.
AST06

Cost to the agent

Instructions loaded
≈7,760 tok
5 files · 80 in bundle
Tool list per prompt
≈3,676 tok
18 tools · 204 per tool
Handshake
441 ms
launch to tools/list, stdio
Undocumented tools
0
description under 20 chars · avg 360

Tokens estimated at four characters each. The tool list is re-read by the agent on every turn, so it is the number to compare between servers that do the same job: ranked across the index.

Known vulnerabilities

No advisories in OSV for symbols-mcp@3.14.786 as of 18 h ago. This is the package itself, not its dependencies.

What would make this Verified

19 points to 80 · weights: behavior 45%, static 25%, prompt 15%, publisher 15%
  1. Text that tells the agent to ignore its instructions≈ +9 on the totalDelete phrases like "ignore previous instructions", "override", "bypass" from SKILL.md and tool descriptions. Describe what the skill does; do not instruct the agent about other rules.
  2. Unclaimed publisher≈ +8 on the totalClaim pypi: a verified publisher gets +50 on publisher trust. It is free and usually instant with GitHub sign-in.
  3. Access to sensitive file paths≈ +4 on the totalStop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
  4. URL with embedded credentials≈ +4 on the totalStop enumerating or reading credential locations; take secrets from a declared environment variable the user sets on purpose.
  5. Auto-update or dynamic fetch configuration≈ +4 on the totalPin versions and sources; vendor anything fetched at install time.
  6. document.cookie access≈ +4 on the totalReplace encoded or assembled strings with plain literals and remove undeclared network calls.
Publish a new version to PyPI, then queue it here. Not your skill? Claim the handle to get verdict changes by email.Claim pypiPropose a patch to the maintainer