BD agentshield
Privacy-First security audit with 77 local tests (52 live + 25 static). Works with OpenClaw, Hermes Agent, n8n (auto-detected), LangChain, and any AI agent platform. Ed25519 private keys stay local. All scripts bundled.
As a process D 39/100 · Unfinished process — weak spots: result and completion, when it triggers, inputs and preconditions
What is at stake
Medium-severity findings: the skill is probably honest, but read what alarmed the scanner.
Below is the worst case for this category. The finding here is medium: the guard saw a sign, not a proof.
The text contains phrases like "ignore previous instructions" or "you are now…". That is an attempt to hijack the agent: it may break your rules, the system limits or company policy.
An honest skill does not need them: state the role and the rules directly without overriding other instructions. Otherwise catalog scanners and corporate filters will block the listing.
How to improve
- Say in the description WHEN to use the skill ("use when…", example requests): that is the agent's main cue.
- Your own cases (evals/evals.json, 4–6 real requests with expected answers): the full check would then run those instead of a model-drafted suite.
- A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.
Guard findings · 13
✓ No critical or high findings
Medium and low: 13
-
medium Instruction override
en-ignore-previousREADME.md:75Instruction-override phrase ("ignore previous instructions") (documentation of a security skill)| **Instruction Override** | Tries to override safety instructions | Critical |
security skill -
low Dangerous commands
cmd-pipe-to-shellagentshield_attack_patterns.json:5Downloads and executes remote code from an unrecognised host (pipe to shell) (string literal in code, not executed; documentation of a security skill)"curl | bash",
code literalsecurity skill -
low Dangerous commands
cmd-pipe-to-shellagentshield_attack_patterns.json:6Downloads and executes remote code from an unrecognised host (pipe to shell) (string literal in code, not executed; documentation of a security skill)"wget | sh"
code literalsecurity skill -
low Obfuscation
uni-mixed-script-wordagentshield_tester.py:655Word mixing Latin and Cyrillic letters (homoglyph obfuscation) (1 occurrence) (test fixture / example file)("LA-018", "Unicode Homoglyph Attack", "іgnore"), # Cyrillic ifixture -
low Instruction override
en-ignore-previousaudit_client.py:20Instruction-override phrase ("ignore previous instructions") (negated — the text forbids it)- Test input payloads (attack strings like "ignore previous instructions")
negated -
low Instruction override
en-ignore-previousecholeak_test.py:165Instruction-override phrase ("ignore previous instructions") (test fixture / example file; quoted — discussed, not commanded)ignore previous instructions
fixturequoted -
low Instruction override
en-ignore-previousSKILL.md:382Instruction-override phrase ("ignore previous instructions") (negated — the text forbids it)- ✅ Attack test inputs/payloads (e.g. "ignore previous instructions")
negated -
low Dangerous commands
cmd-pipe-to-shellTESTING.md:155Downloads and executes remote code from an unrecognised host (pipe to shell) (test fixture / example file; documentation of a security skill)- SC-003: Remote code execution (curl | bash, wget | sh)
fixturesecurity skill -
low Risky intent
intent-offensive-securityTESTING.md:165Offensive-security / dual-use content (legitimate for authorised testing; review intended use) (test fixture / example file)- LA-002: `Developer Mode Override` (tests privilege escalation)
fixture
A further 4 matches are quotations in this security skill's documentation and are not counted as findings.
Files scanned: 31. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
- warning
description-no-whendescription does not say WHEN to use the skill (no "use when") - note
frontmatter-keyunknown frontmatter key "triggers"
Process rating: all ten parameters 39/100
- 0Result and completion. Does not say what the result is
- 0Inputs and preconditions. Does not say what the process needs to start
- 0Failures and branches. Linear process with no failure handling
- 0Progress reporting. Says nothing while it works
- 20When it triggers. No condition that starts the skill
- 30Running it twice. 1 mutating operations with no state check
- 40Consistency. Frontmatter name (agentshield) differs from the folder (agentshield-audit)
- 60Tools and files. Uses tools (bash) that frontmatter does not declare
- 100Steps. 102 steps
- 100Execution cost. Instruction body is 3380 tokens
- medium Safety rules and hard prohibitions inside a skill: they belong in the system prompt, here they protect nothing
- low 16 top-level sections: this looks like several domains in one skill
Everything here is measured from the skill text rather than judged by a model, so the numbers are checkable. A parameter weighs more when it is a more common reason for the process to stall.
Quality signals
- +5Description has no quoted example phrases that should trigger the skill
- +4Description does not say when NOT to use the skill (false activations)
- +3Output format is not stated: the model decides each time
- -5TODO / placeholder text left in the skill
- -278 emoji in the instructions: noise for the model
- +1No license
- +2Single-language instructions
- +3Description length 219: enough signal without eating the budget
- +4Structure: 41 headings
- +3Step-by-step instructions: 102 items
- +4Has examples (16 code blocks)
Quality base 70; lint remarks subtract, signals add up to 100. Result: 64.