Back to skill

Security audit

Tristan RFQ overseer

Security checks for vulnerabilities and agentic risk

Overview

This RFQ skill is purpose-aligned and disclosed, but users should review its business-data and pricing effects before relying on it operationally.

Install only after connecting it to the intended Obsidian vault, Telegram channel, and email account. Review generated prices, supplier rankings, and draft emails manually, and keep explicit send confirmation enabled. Use pinned dependencies or an isolated environment for publishing/validation tools.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (4)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/pricing_model.py:72
Finding

Unvalidated pricing inputs can produce economically invalid quotations

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/pricing_strategies.py:47
Finding

Pricing strategy parameters lack domain and range validation

Content
View full analysis
dict: """Position the quote relative to observed competitor/market prices. target_percentile: 0.0 = match the lowest observed price, 1.0 = match the highest, 0.5 = market median-ish. Never suggests going below the cost-plus baseline — that would be selling at a loss regardless of what the market is doing. """ if not market_prices: raise ValueError("At least one market price is required.") lo, hi = min(market_prices), max(market_prices) avg = round(mean(market_prices), 2) raw_suggested = lo + (hi - lo) * target_percentile suggested = round(max(raw_suggested, baseline_total), 2) ``` ```python def value_based(baseline_total: float, perceived_value: float, value_capture_rate: float = 0.3) -> dict: """Price toward a share of the value delivered to the client, never below the cost-plus floor.""" if perceived_value <= 0: raise ValueError("perceived_value must be positive.") raw_suggested = round(perceived_value * value_capture_rate, 2) suggested = max(raw_suggested, baseline_total) ``` ```python def target_costing(target_price: float, desired_margin: float, overhead_rate: float) -> dict: """Work backward from a market-driven target price to the maximum allowable cost base.""" if desired_margin >= 1: raise ValueError("desired_margin must be a fraction < 1, e.g. 0.18 for 18%.") max_cost_base = round(target_price / (1 + desired_margin), 2) max_subtotal = round(max_cost_base / (1 + overhead_rate), 2) ``` ```python def index_escalation(base_price: float, index_rate: float, periods: int, period_label: str = "year") ...[truncated 2156 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/compare_quotes.py:54
Finding

Supplier quote ranking accepts invalid values and unsafe weighting parameters

Content
View full analysis
1e-6: print(f"Warning: weights sum to {weight_sum}, not 1.0 — scores will be scaled accordingly.", file=sys.stderr) quotes = load_quotes(args.quotes_dir) if not quotes: print("No valid supplier quote files found.", file=sys.stderr) sys.exit(1) ranked = rank_ ...[truncated 1888 chars]
Remediation
View remediation

T08 · Insecure Dependencies

Note
Location
README.md:111
Finding

Installation guidance uses unpinned global dependencies and bypasses system package protections

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (6)

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The declared description promises a broad, end-to-end RFQ workflow with communications intake, storage, pricing, comparison, and drafting behavior. The actual code chunk only implements one narrow stage: comparing/ranking supplier quotes from existing JSON files in a directory. It does not process inbound emails, Telegram messages, wake phrases, vault storage, or draft outbound communications. While supplier quote comparison is part of the declared pipeline, the code does not substantiate most of the stated functionality, so the description materially overstates the implemented behavior.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The declared description presents a broad, end-to-end RFQ automation skill with communication-channel triggers and multiple stages: intake from email/Telegram, Obsidian storage, pricing, supplier comparison, and drafting outbound confirmations. The supplied code implements only one narrow stage: pricing calculation for an existing local markdown RFQ note, with optional write-back to a Pricing section. It does not interact with email, Telegram, wake phrases, supplier quotes, outbound drafting, or orchestration logic. While the pricing portion is consistent with part of the description, the actual code is materially narrower than the declared primary purpose, so this is a mismatch.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The declared description presents a broad RFQ automation skill with communications intake, storage/orchestration, quote comparison, and outbound drafting. The supplied code chunk is narrowly focused on pricing strategy calculations and markdown report generation for an existing note file. It accepts command-line subcommands, reads a local note path, computes baseline pricing from parsed line items, and optionally writes a pricing comparison section back to that file. There is no code for email/Telegram integration, trigger handling, wake phrase detection, supplier quote ingestion/comparison workflow, or drafting/sending confirmations. While pricing calculation is one subset of the declared purpose, the overall description materially overstates the implemented behavior, so this is a clear mismatch.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

There is a clear description-behavior mismatch. The declared purpose is an operational procurement/RFQ automation skill with messaging and email triggers plus business workflow functions. The actual code does none of that. It is a developer utility for validating a package prior to publishing: compiling Python files, checking regex compatibility, validating YAML frontmatter, checking file references, and finding build artifacts. This is not a supporting implementation detail of the RFQ pipeline; it is a separate packaging/QA function with a materially different primary purpose.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
89% confidence
Finding

The skill declares file read/write capable behavior but does not define any explicit tool scope such as permissions or allowed-tools. In an agent environment, this increases the chance of over-broad filesystem access, making unintended reads or writes to vault data or adjacent files more likely if the runtime grants defaults that are too permissive.

Content

No source excerpt is available for this finding.

Missing User Warnings

Low
Category
Not specified by scanner
Confidence
83% confidence
Finding

This code rewrites the user-supplied note file in place by replacing or appending a markdown section. Although the operation is gated behind the --write flag, there is no confirmation prompt and no comment/docstring near the write path warning that the original file will be modified.

Content

No source excerpt is available for this finding.