Back to skill

Security audit

smb-sales-boost

Security checks for vulnerabilities and agentic risk

Overview

This skill is high-impact but transparent: it can spend credits, charge money, delete settings, email leads, and send lead data to third parties, but those actions are disclosed and gated by confirmation.

Install only if you intend to let an agent operate your SMB Sales Boost account. Keep the API key in the environment, review every action before approving --confirm, set caps on lead searches/exports, and be especially careful with purchases, auto-billing, email schedules, deletes, and CRM/webhook destinations.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (38)

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · REFERENCE.md (reported line 29)May include surrounding context.

md
| `POST /lead-export-history/re-export` | Rebuild a file from stored copies | free | export |
| `POST /lead-export-history/refresh-and-export` | Refresh, then rebuild a file | free; sends `lead.updated` like refresh | export |
| `GET /export-formats`, `GET /export-formats/{id}` | List or get export formats | | |
| `POST /export-formats`, `PATCH /export-formats/{id}`, `DELETE /export-formats/{id}` | Create, update, delete | | |
| `POST /export-formats/{id}/set-default` | Make a format the default | | |
| `GET /export-blacklist`, `POST /export-blacklist`, `DELETE /export-blacklist/{id}` | Domains to keep out of exports | | |
| `GET /filter-presets`, `POST /filter-presets`, `DELETE /filter-presets/{id}` | Saved searches for email schedules | | |

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · REFERENCE.md (reported line 31)May include surrounding context.

md
| `GET /export-formats`, `GET /export-formats/{id}` | List or get export formats | | |
| `POST /export-formats`, `PATCH /export-formats/{id}`, `DELETE /export-formats/{id}` | Create, update, delete | | |
| `POST /export-formats/{id}/set-default` | Make a format the default | | |
| `GET /export-blacklist`, `POST /export-blacklist`, `DELETE /export-blacklist/{id}` | Domains to keep out of exports | | |
| `GET /filter-presets`, `POST /filter-presets`, `DELETE /filter-presets/{id}` | Saved searches for email schedules | | |
| `GET /keyword-lists`, `POST /keyword-lists`, `PUT /keyword-lists/{id}`, `DELETE /keyword-lists/{id}` | Keyword lists | | |
| `GET /email-schedules`, `POST /email-schedules`, `PATCH /email-schedules/{id}`, `DELETE /email-schedules/{id}` | Scheduled lead emails | credits, sends email (an active schedule, the default on create, sends its first email within about 15 minutes) | |

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · REFERENCE.md (reported line 216)May include surrounding context.

md
| `GET /export-formats`, `GET /export-formats/{id}` | List or get export formats | | |
| `POST /export-formats`, `PATCH /export-formats/{id}`, `DELETE /export-formats/{id}` | Create, update, delete | | |
| `POST /export-formats/{id}/set-default` | Make a format the default | | |
| `GET /export-blacklist`, `POST /export-blacklist`, `DELETE /export-blacklist/{id}` | Domains to keep out of exports | | |
| `GET /filter-presets`, `POST /filter-presets`, `DELETE /filter-presets/{id}` | Saved searches for email schedules | | |
| `GET /keyword-lists`, `POST /keyword-lists`, `PUT /keyword-lists/{id}`, `DELETE /keyword-lists/{id}` | Keyword lists | | |
| `GET /email-schedules`, `POST /email-schedules`, `PATCH /email-schedules/{id}`, `DELETE /email-schedules/{id}` | Scheduled lead emails | credits, sends email (an active schedule, the default on create, sends its first email within about 15 minutes) | |

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · REFERENCE.md (reported line 32)May include surrounding context.

md
| `POST /export-formats`, `PATCH /export-formats/{id}`, `DELETE /export-formats/{id}` | Create, update, delete | | |
| `POST /export-formats/{id}/set-default` | Make a format the default | | |
| `GET /export-blacklist`, `POST /export-blacklist`, `DELETE /export-blacklist/{id}` | Domains to keep out of exports | | |
| `GET /filter-presets`, `POST /filter-presets`, `DELETE /filter-presets/{id}` | Saved searches for email schedules | | |
| `GET /keyword-lists`, `POST /keyword-lists`, `PUT /keyword-lists/{id}`, `DELETE /keyword-lists/{id}` | Keyword lists | | |
| `GET /email-schedules`, `POST /email-schedules`, `PATCH /email-schedules/{id}`, `DELETE /email-schedules/{id}` | Scheduled lead emails | credits, sends email (an active schedule, the default on create, sends its first email within about 15 minutes) | |
| `POST /email-schedules/{id}/trigger` | Send a schedule now | credits, sends email | export |

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · REFERENCE.md (reported line 33)May include surrounding context.

md
| `POST /export-formats/{id}/set-default` | Make a format the default | | |
| `GET /export-blacklist`, `POST /export-blacklist`, `DELETE /export-blacklist/{id}` | Domains to keep out of exports | | |
| `GET /filter-presets`, `POST /filter-presets`, `DELETE /filter-presets/{id}` | Saved searches for email schedules | | |
| `GET /keyword-lists`, `POST /keyword-lists`, `PUT /keyword-lists/{id}`, `DELETE /keyword-lists/{id}` | Keyword lists | | |
| `GET /email-schedules`, `POST /email-schedules`, `PATCH /email-schedules/{id}`, `DELETE /email-schedules/{id}` | Scheduled lead emails | credits, sends email (an active schedule, the default on create, sends its first email within about 15 minutes) | |
| `POST /email-schedules/{id}/trigger` | Send a schedule now | credits, sends email | export |
| `POST /ai/suggest-categories` | Suggest customer categories | | AI |

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · REFERENCE.md (reported line 34)May include surrounding context.

md
| `GET /export-blacklist`, `POST /export-blacklist`, `DELETE /export-blacklist/{id}` | Domains to keep out of exports | | |
| `GET /filter-presets`, `POST /filter-presets`, `DELETE /filter-presets/{id}` | Saved searches for email schedules | | |
| `GET /keyword-lists`, `POST /keyword-lists`, `PUT /keyword-lists/{id}`, `DELETE /keyword-lists/{id}` | Keyword lists | | |
| `GET /email-schedules`, `POST /email-schedules`, `PATCH /email-schedules/{id}`, `DELETE /email-schedules/{id}` | Scheduled lead emails | credits, sends email (an active schedule, the default on create, sends its first email within about 15 minutes) | |
| `POST /email-schedules/{id}/trigger` | Send a schedule now | credits, sends email | export |
| `POST /ai/suggest-categories` | Suggest customer categories | | AI |
| `POST /ai/generate-keywords` | Delete all keyword lists and regenerate them | | AI |

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · REFERENCE.md (reported line 56)May include surrounding context.

md
| `GET /integrations`, `GET /integrations/{id}` | List integrations, get one | | integrations |
| `POST /integrations/webhook` | Create a webhook integration (returns the signing secret once) | sends data out (from then on, automatically) | integrations |
| `PATCH /integrations/{id}` | Rename, pause, re-enable, change events, destination or header `apiKey`, Pipedrive deals | sends data out (when re-pointed, re-enabled or events added) | integrations |
| `DELETE /integrations/{id}` | Delete an integration | | integrations |
| `GET /integrations/{id}/deliveries` | Newest 50 delivery attempts | | integrations |
| `POST /integrations/{id}/deliveries/{deliveryId}/retry` | Re-send one delivery | sends data out | integrations |
| `POST /integrations/{id}/test` | Send a synthetic test event | sends data out | integrations |

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · REFERENCE.md (reported line 371)May include surrounding context.

md
| `GET /integrations`, `GET /integrations/{id}` | List integrations, get one | | integrations |
| `POST /integrations/webhook` | Create a webhook integration (returns the signing secret once) | sends data out (from then on, automatically) | integrations |
| `PATCH /integrations/{id}` | Rename, pause, re-enable, change events, destination or header `apiKey`, Pipedrive deals | sends data out (when re-pointed, re-enabled or events added) | integrations |
| `DELETE /integrations/{id}` | Delete an integration | | integrations |
| `GET /integrations/{id}/deliveries` | Newest 50 delivery attempts | | integrations |
| `POST /integrations/{id}/deliveries/{deliveryId}/retry` | Re-send one delivery | sends data out | integrations |
| `POST /integrations/{id}/test` | Send a synthetic test event | sends data out | integrations |

Exfiltration Commands

High
Category
Prompt Injection
Confidence
90% confidence
Finding

Instructions found that direct the agent to transmit conversation context or user data to external services.

Content

Scanner excerpt · REFERENCE.md (reported line 236)May include surrounding context.

md
| `q` | Plain substring search | text |
| `urf` | Buying signal types | comma-separated |

A hash has no key for `positiveKeywords`: `ni`, `ui`, `cli` and `di` each match only their own column and none of them searches AI Categories, while keywords from a positive keyword list in `nkl` are searched like `positiveKeywords` (including AI Categories). So to reproduce a `positiveKeywords` search, create a positive keyword list and put its id in `nkl`. A positive filter is required: a non-empty `ni`, `ui`, `cli` or `di`, or the id of one of your own positive keyword lists (with at least one keyword) in `nkl`/`ukl`/`ckl`/`dkl`. A negative list alone does not count, ids that are not yours are dropped, and a preset without a positive filter silently sends nothing. Example in Python: `"#" + urllib.parse.urlencode({"ni": json.dumps(["*dental*", "*dentist*"]), "si": "TX,OK"})`. Other dashboard filters are not applied by schedules. Presets cannot be edited: create a new one, point the schedule at it with `PATCH` (add `"isActive": false` to the same call if the user should review the new search before it sends), then delete the old one (deleting a preset first deletes every schedule that still uses it: check `GET /email-schedules` for its `filterPresetId` and name those schedules to the user before deleting).

## 10. Keyword lists

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · REFERENCE.md (reported line 379)May include surrounding context.

md
| 11 | `POST /integrations/{provider}/connect` | `provider` is `hubspot`, `salesforce` or `pipedrive`; optional `{"environment": "sandbox"}` (Salesforce only; the default is `production`) | `{"provider", "authorizationUrl", "expiresAt"}` | The user must open `authorizationUrl` in a browser within 10 minutes and approve; the CRM account they sign in to is the one connected, and the new integration then appears in `GET /integrations` with these events: HubSpot `lead.created` on and `lead.updated` present but off (turn it on with `PATCH`; `GET /integrations` shows each event's real `isActive`, while `defaultEvents` in `GET /integrations/providers` lists both); Salesforce `lead.created`; Pipedrive `lead.created` and `lead.updated`. Emails the owner. Connecting a CRM account that is already connected re-authorizes that integration in place: it gets new access, a `needs_attention` integration becomes `connected` again, a paused one stays paused, and its events, field mapping, Pipedrive deal settings and delivery history are kept (no new integration, and the owner is emailed only when it brings a `needs_attention` integration back). Use this when a CRM integration is `needs_attention` because its access was removed or expired; the user must sign in to the same CRM account, and a Salesforce sandbox integration needs `{"environment": "sandbox"}`. Signing in to a different account of the same CRM connects it as a new integration. If HubSpot does not confirm which account the user signed in to, nothing is connected or changed ("We could not confirm which HubSpot account you signed in to, so nothing was connected or changed. Please try connecting again in a minute."); ask for a new link a minute later. If the API key that asked for the link is revoked before the user approves, the connection fails with `key_revoked` and nothing is created. `--confirm`. Safe to repeat (a new link). `400 bad_request` only for a wrong `environment` (sent for a provider other than Salesforce, or no
...[truncated 26 chars]

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 15)May include surrounding context.

md
o rule, parameter or example is missed. Full parameter and field tables are in `REFERENCE.md` in this folder.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 26)May include surrounding context.

md
o rule, parameter or example is missed. Full parameter and field tables are in `REFERENCE.md` in this folder.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 53)May include surrounding context.

md
o rule, parameter or example is missed. Full parameter and field tables are in `REFERENCE.md` in this folder.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 120)May include surrounding context.

md
o rule, parameter or example is missed. Full parameter and field tables are in `REFERENCE.md` in this folder.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 131)May include surrounding context.

md
o rule, parameter or example is missed. Full parameter and field tables are in `REFERENCE.md` in this folder.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 146)May include surrounding context.

md
o rule, parameter or example is missed. Full parameter and field tables are in `REFERENCE.md` in this folder.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 183)May include surrounding context.

md
o rule, parameter or example is missed. Full parameter and field tables are in `REFERENCE.md` in this folder.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 206)May include surrounding context.

md
o rule, parameter or example is missed. Full parameter and field tables are in `REFERENCE.md` in this folder.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 210)May include surrounding context.

md
o rule, parameter or example is missed. Full parameter and field tables are in `REFERENCE.md` in this folder.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 220)May include surrounding context.

md
o rule, parameter or example is missed. Full parameter and field tables are in `REFERENCE.md` in this folder.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 243)May include surrounding context.

md
o rule, parameter or example is missed. Full parameter and field tables are in `REFERENCE.md` in this folder.

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 26)May include surrounding context.

md
*Reference files:** `REFERENCE.md` (every parameter, field, status and error), `openapi.json` (the published OpenAPI 3.1 spec; where it differs from this skill,

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · SKILL.md (reported line 215)May include surrounding context.

md
## AI helpers

- `POST /ai/suggest-categories` with `companyName`, `companyDescription` and `productService` (optional `companyWebsite`, `excludeCategories`) returns 4-12 suggested customer categories for the user's business. Nothing is saved or charged. With a website to analyse it can take a while; the script follows the call until the suggestions are ready. Turn the suggestions into keywords yourself (for example "Dentists" becomes `["*dental*","*dentist*"]`).
- `POST /ai/generate-keywords` **deletes every existing keyword list** and regenerates lists from the account's target categories, which can only be set in the dashboard. If `GET /me` shows an empty `targetCategories`, it deletes the lists and generates nothing. Filter presets that point at keyword lists by id (`nkl`, `ukl`, `ckl` or `dkl` in the hash, which dashboard-saved presets also use) silently lose those keywords and exclusions when the lists are deleted, and the same applies to `DELETE /keyword-lists/{id}`: their email schedules can stop sending, or start emailing a broader set of leads and spending credits on them. Check `GET /filter-presets` first and pause any affected schedules. Only use it when the user explicitly wants all lists replaced (needs `--confirm`). Check progress with `GET /ai/keyword-status`.
- `POST /ai/auto-refine/enable` / `disable` with `{"listId": 42}` turn on or off the AI refinement of a keyword list and its paired list; `GET /ai/auto-refine/status?listId=42` shows progress. Enabling resets the list's score history, and the refinement rewrites the list's keywords, which changes what any email schedule using the list sends (pause those schedules first if the user wants to review). A `200` does not guarantee the run started: if `autoRefineEnabled` is true but `refinementStatus` stays `null`, call enable again a few minutes later.

## Integrations: sending leads to a CRM or automation tool

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

The skill includes capabilities to delete integrations, re-point destinations, push leads, retry deliveries, and send tests to external endpoints, all of which can transmit or destroy sensitive lead-routing configuration. While consent requirements are documented, the skill text also states some high-risk changes may partially apply before a 500 error, creating a realistic risk of duplicate transmissions or unintended destructive changes if the agent or operator mishandles retries.

Content

Scanner excerpt · SKILL.md (reported line 227)May include surrounding context.

md
- **Webhook integration:** `POST /integrations/webhook` with `provider` (`zapier`, `n8n`, `make`, `pipedream`, `clay` or `generic_webhook`), `targetUrl` (a public `https` URL), `events` and optional `displayName` (and `apiKey` for Clay, passed with `--body-file`). It needs `--confirm`. The response has the integration and a `signingSecret` that is never shown again: give it to the user so their endpoint can verify signatures, and do not log or store it. If the call fails or times out, the integration may have been created: check `GET /integrations` before trying again. Repeating the same provider and `targetUrl` returns `409 integration_exists`; a different provider or a changed URL (even a trailing slash) creates a second integration with its own secret. The signing secret cannot be shown again, so if the integration exists but the user did not get its `signingSecret`: read it with `GET /integrations/{id}`, show the user its provider, destination URLs and events, and with their yes delete it and create it again with the same values (both calls need `--confirm`; the new integration gets a new secret, and events that went to different URLs are restored with a `PATCH` of `subscriptions`).
- **CRM connection:** `POST /integrations/{provider}/connect` (`hubspot`, `salesforce` with optional `{"environment": "sandbox"}`, or `pipedrive`; needs `--confirm`) returns an `authorizationUrl`. The user opens it in a browser within 10 minutes and approves; the CRM account they sign in to is the one connected. Then the integration appears in `GET /integrations` with these events: HubSpot `lead.created` (its `lead.updated` subscription exists but starts switched off), Salesforce `lead.created`, Pipedrive `lead.created` and `lead.updated`. HubSpot creates or updates Contacts matched by email (a lead without an email creates no Contact) and Companies matched by website domain, or by exact company name when the lead has no website of its own (none, or a social or listing page such as F
...[truncated 26 chars]

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
80% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · openapi.json (reported line 1863)May include surrounding context.

json
}
          },
          "400": {
            "description": "validation_error (details lists the problems): a scope that belongs to another CRM (for example 'Scope(s) lead do not apply to HubSpot integrations. HubSpot takes: contact, company.'), or no scope at all (send at least one scope; use DELETE /integrations/{id}/field-mapping to reset). Or bad_request (invalid id, or the integration is not a HubSpot, Salesforce or Pipedrive integration).",
            "content": {
              "application/json": {
                "schema": {