Back to skill

Security audit

GTM Coach

Security checks for vulnerabilities and agentic risk

Overview

The skill is a disclosed GTM coaching tool, but it gives itself persistent agent-install and automatic remote self-update authority without user confirmation.

Review before installing. Only use this skill if you are comfortable with it writing local .gtm files, storing an nmagt token in .gtm/secrets.md, installing itself into your agent skills folder, and automatically replacing that installed copy from nichestarter.ai on new sessions. Keep .gtm/ gitignored and do not use the dashboard share buttons for sensitive business details.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • System Prompt LeakageDirect Leakage, Indirect Extraction, Tool-Based Exfiltration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Findings (22)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 90)May include surrounding context.

md
**current** host’s skills folder. Do not copy only `SKILL.md`. Do not install

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 116)May include surrounding context.

md
**current** host’s skills folder. Do not copy only `SKILL.md`. Do not install

Direct Prompt Extraction

High
Category
System Prompt Leakage
Confidence
85% confidence
Finding

Skill contains instructions that could directly expose system prompts, internal rules, or hidden instructions to users or external parties.

Content

Scanner excerpt · channels/newsletter-email.md (reported line 16)May include surrounding context.

md
## Lifecycle loop

Trigger email from real user state: welcome/first value, incomplete activation, achieved outcome, relevant update, return prompt, or renewal. One email should help one next action. Respect consent, unsubscribe, frequency, and applicable law.

## Measure

Anti-Refusal Statement

High
Category
Anti-Refusal
Confidence
85% confidence
Finding

Skill instructs the agent to omit warnings, disclaimers, or ethical commentary. Stripping safety caveats hides risk from the user and is a common jailbreak preamble.

Content

Scanner excerpt · channels/x.md (reported line 26)May include surrounding context.

md
Primary: qualified replies/DMs → profile-to-site action → conversations/activation/payment. Diagnose impressions, engagement, profile visits, and link clicks.

Continue a recurring theme that attracts the right people. Change format/hook if the right audience sees but does not stop; change topic/offer if engagement comes only from peers outside the ICP. Do not judge from one post.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger description includes broad natural-language phrases such as references to traffic, growth spike, this week, or what to do in go-to-market, which can match ordinary user conversation and cause the skill to activate unexpectedly. In this skill’s context, unexpected activation is more concerning because the changelog shows persistent installation behavior and authority to write into workspace and agent skill folders, increasing the chance of unintended file changes or steering of the session.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The invocation phrases are broad enough to activate on ordinary marketing or growth discussions, which can cause the agent to apply this skill unexpectedly. Because the skill then instructs the agent to persist files, call remote services, and manage state, accidental invocation can trigger unintended side effects beyond simple chat behavior.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill explicitly directs the agent to copy or symlink itself into host skill directories and to write strategy and secrets into workspace files, but does not require clear upfront user consent for those filesystem changes. In practice this enables persistence and local modification of agent configuration, which increases the blast radius of accidental invocation or prompt injection-driven activation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill mandates an automatic remote update check and immediate auto-application from a network source at each new session, without requiring user confirmation or providing an integrity-verification step. This creates a supply-chain and privacy risk: remote content can change local behavior and overwrite the persisted skill bundle, while session metadata and tokens may influence network requests.

Content

No source excerpt is available for this finding.

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · SKILL.md (reported line 328)May include surrounding context.

md
Confirm (store the evidence value exactly as in `plan.json`):

| Field              | Options                                                                                                                                                               |
| ------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Project type       | `MICRO_SAAS` · `SERVICE_FREELANCER` · `NEWSLETTER_MEDIA` · `DIGITAL_PRODUCT` · `COMMUNITY` · `CREATOR_ARTIST` · `BRAND` · `EDUCATION_COHORT` · `PLATFORM_MARKETPLACE` |
| Period             | `validate` · `build` · `cold_start` · `launch` · `promote` · `monitor`                                                                                                |

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · SKILL.md (reported line 331)May include surrounding context.

md
| Field              | Options                                                                                                                                                               |
| ------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Project type       | `MICRO_SAAS` · `SERVICE_FREELANCER` · `NEWSLETTER_MEDIA` · `DIGITAL_PRODUCT` · `COMMUNITY` · `CREATOR_ARTIST` · `BRAND` · `EDUCATION_COHORT` · `PLATFORM_MARKETPLACE` |
| Period             | `validate` · `build` · `cold_start` · `launch` · `promote` · `monitor`                                                                                                |
| Readiness          | not ready · first value deliverable this week · live · repeat customers/audience                                                                                      |
| Evidence           | `none` · `anecdote` · `pains_scored` · `waitlist` · `paying`                                                                                                          |
| Audience and offer | named ICP + job/pain + promise + CTA                                                                                                                                  |

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · SKILL.md (reported line 332)May include surrounding context.

md
| ------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Project type       | `MICRO_SAAS` · `SERVICE_FREELANCER` · `NEWSLETTER_MEDIA` · `DIGITAL_PRODUCT` · `COMMUNITY` · `CREATOR_ARTIST` · `BRAND` · `EDUCATION_COHORT` · `PLATFORM_MARKETPLACE` |
| Period             | `validate` · `build` · `cold_start` · `launch` · `promote` · `monitor`                                                                                                |
| Readiness          | not ready · first value deliverable this week · live · repeat customers/audience                                                                                      |
| Evidence           | `none` · `anecdote` · `pains_scored` · `waitlist` · `paying`                                                                                                          |
| Audience and offer | named ICP + job/pain + promise + CTA                                                                                                                                  |
| Primary channel    | one id from `channels/index.md`                                                                                                                                       |

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · SKILL.md (reported line 333)May include surrounding context.

md
| Project type       | `MICRO_SAAS` · `SERVICE_FREELANCER` · `NEWSLETTER_MEDIA` · `DIGITAL_PRODUCT` · `COMMUNITY` · `CREATOR_ARTIST` · `BRAND` · `EDUCATION_COHORT` · `PLATFORM_MARKETPLACE` |
| Period             | `validate` · `build` · `cold_start` · `launch` · `promote` · `monitor`                                                                                                |
| Readiness          | not ready · first value deliverable this week · live · repeat customers/audience                                                                                      |
| Evidence           | `none` · `anecdote` · `pains_scored` · `waitlist` · `paying`                                                                                                          |
| Audience and offer | named ICP + job/pain + promise + CTA                                                                                                                                  |
| Primary channel    | one id from `channels/index.md`                                                                                                                                       |
| Optional campaign  | none, or one explicitly approved channel/campaign                                                                                                                     |

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · SKILL.md (reported line 334)May include surrounding context.

md
| Period             | `validate` · `build` · `cold_start` · `launch` · `promote` · `monitor`                                                                                                |
| Readiness          | not ready · first value deliverable this week · live · repeat customers/audience                                                                                      |
| Evidence           | `none` · `anecdote` · `pains_scored` · `waitlist` · `paying`                                                                                                          |
| Audience and offer | named ICP + job/pain + promise + CTA                                                                                                                                  |
| Primary channel    | one id from `channels/index.md`                                                                                                                                       |
| Optional campaign  | none, or one explicitly approved channel/campaign                                                                                                                     |

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · SKILL.md (reported line 335)May include surrounding context.

md
| Readiness          | not ready · first value deliverable this week · live · repeat customers/audience                                                                                      |
| Evidence           | `none` · `anecdote` · `pains_scored` · `waitlist` · `paying`                                                                                                          |
| Audience and offer | named ICP + job/pain + promise + CTA                                                                                                                                  |
| Primary channel    | one id from `channels/index.md`                                                                                                                                       |
| Optional campaign  | none, or one explicitly approved channel/campaign                                                                                                                     |

`pains_scored` means observed product use or scored pains — not a Nichestarter

Whitespace Padding

Medium
Category
Prompt Injection
Confidence
70% confidence
Finding

Large whitespace padding was detected (a block of blank lines or a long run of spaces). This can push injected instructions below or to the right of the visible area so a human reviewer never sees them while the agent still reads them. Manual review of the hidden content is recommended.

Content

Scanner excerpt · SKILL.md (reported line 336)May include surrounding context.

md
| Evidence           | `none` · `anecdote` · `pains_scored` · `waitlist` · `paying`                                                                                                          |
| Audience and offer | named ICP + job/pain + promise + CTA                                                                                                                                  |
| Primary channel    | one id from `channels/index.md`                                                                                                                                       |
| Optional campaign  | none, or one explicitly approved channel/campaign                                                                                                                     |

`pains_scored` means observed product use or scored pains — not a Nichestarter
dump. If they have retained/referring customers, store `paying` and put the

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The template embeds social-sharing and post-drafting workflows directly into the dashboard, including Facebook, X, LinkedIn, and Threads helpers. In this skill's stated context, that is a policy-boundary violation: GTM Coach is supposed to keep strategy local and explicitly avoid handling social post/comment workflows, so these features expand the skill's operational scope and can cause unintended disclosure of workspace-derived content to external platforms.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
87% confidence
Finding

The code adds multi-platform publishing helpers that are not necessary for rendering a local GTM dashboard and that encourage pushing generated content to third-party services. Even though it uses share URLs rather than direct API posting, it still creates an exfiltration path for strategy content and increases the chance that sensitive or premature business information is copied or shared outside the local workspace.

Content

No source excerpt is available for this finding.

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · SKILL.md (reported line 616)May include surrounding context.

md
**Not a v1 launch pillar.** Only if `GET /status` → `match.marketplace.enabled` is true.

Then follow  
https://www.nichestarter.ai/skills/nichematch-agent-onboarding/SKILL.md  
with the **same** `nmagt_` key.

Rules to keep in this chat even before you open that skill:

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
80% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · playbook.md (reported line 280)May include surrounding context.

md
**Not a v1 launch pillar.** Only if `GET /status` → `match.marketplace.enabled` is true.

Then follow  
https://www.nichestarter.ai/skills/nichematch-agent-onboarding/SKILL.md  
with the **same** `nmagt_` key.

Rules to keep in this chat even before you open that skill:

Skill Enumeration

Medium
Category
Agent Snooping
Confidence
85% confidence
Finding

Skill enumerates or reads other installed skills. Access to other skills' SKILL.md files or the skills directory reveals prompt instructions, capabilities, and secrets that should be invisible to peer skills.

Content

Scanner excerpt · reference.md (reported line 324)May include surrounding context.

.nichematch/

text

Do **not** gitignore the agent skills folder. It has no secrets. Optional: commit project-level `gtm-coach/` under `.cursor/skills`, `.claude/skills`, `.agents/skills`, and/or `.codebuddy/skills` (or symlink one canonical copy) so teammates on that host get natural-language calling. `.gtm/` stays local and is readable by any agent that opens the same workspace.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

The HTML sets lang="en", which hard-codes English as the page language. Under the policy, locale constraints should be optional or clearly justified; this template provides neither a user opt-in nor documentation that the dashboard is intended to be English-only.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

The instruction requires the exact English signature Via gtm_coach by nichestarter.ai, which forces a specific language output. The file does not offer any locale or language opt-in for this user-facing text, so it may violate language choice policy.

Content

No source excerpt is available for this finding.