DC hermes-agent
Expert in building self-improving AI agents with tool use, multi-platform messaging, and a closed learning loop. Proficient in LLM orchestration, tool integration, session management, and agent autonomy.
Expert in building self-improving AI agents with tool use, multi-platform messaging, and a closed learning loop.
As a process C 50/100 · Has gaps — weak spots: result and completion, execution cost, running it twice
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 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.
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 skill tells the agent to hide things from you: not to show errors, not to mention actions, to report differently from what was done. You lose the ability to see what the agent really did.
Transparency beats a smooth answer. If the goal is to hide technical noise, ask the agent to "summarise briefly", not to "not mention".
How to improve
- The SKILL.md body is over 5,000 tokens: move reference detail into references/ and load it when needed.
- 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 · 10
✓ No critical or high findings
Medium and low: 10
-
medium Dangerous commands
cmd-pipe-to-shell-known-hostSKILL.md:129Pipe-to-shell installer from a well-known host (still executes remote code)curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash
-
medium Dangerous commands
cmd-pipe-to-shell-known-hostSKILL.md:150Pipe-to-shell installer from a well-known host (still executes remote code)curl -LsSf https://astral.sh/uv/install.sh | sh
-
medium Exfiltration
exfil-read-secret-filesSKILL.md:997Reads credential / secret files (quoted — discussed, not commanded)- Secret access (`cat ~/.env`, `cat ~/.netrc`)
quoted -
medium Instruction override
en-ignore-previousSKILL.md:1056Instruction-override phrase ("ignore previous instructions") (quoted — discussed, not commanded)- "ignore previous instructions", "system prompt override"
quoted -
medium Instruction override
en-fake-system-promptSKILL.md:1056Fake system prompt injected into content (quoted — discussed, not commanded)- "ignore previous instructions", "system prompt override"
quoted -
medium Concealment
en-hide-from-userSKILL.md:1057Instruction to hide actions from the user (quoted — discussed, not commanded)- "do not tell the user", "act as if you have no restrictions"
quoted -
low Dangerous commands
cmd-privilegeSKILL.md:992Privilege escalation / world-writable permissions (quoted — discussed, not commanded)- Privilege escalation (`chmod 777`, `chown -R root`)
quoted -
low Risky intent
intent-offensive-securitySKILL.md:992Offensive-security / dual-use content (legitimate for authorised testing; review intended use)- Privilege escalation (`chmod 777`, `chown -R root`)
-
low Exfiltration
read-dotenvSKILL.md:997Reads a .env file (quoted — discussed, not commanded)- Secret access (`cat ~/.env`, `cat ~/.netrc`)
quoted -
low Risky intent
intent-offensive-securitySKILL.md:1488Offensive-security / dual-use content (legitimate for authorised testing; review intended use) (detector / deny-list definition)- 106 dangerous command patterns (destructive, privilege escalation, SQL, exfiltration)
detector
Files scanned: 1. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
- warning
body-longSKILL.md body ≈ 18510 tokens (recommended < 5000); move details to references/
Process rating: all ten parameters 50/100
- 0Result and completion. Does not say what the result is
- 10Execution cost. Instruction body is 18510 tokens: crowds the task out of the window
- 30Running it twice. 20 mutating operations with no state check
- 50When it triggers. No condition that starts the skill
- 50Failures and branches. 0 branches, has a failure section
- 60Tools and files. Uses tools (bash, web, python) that frontmatter does not declare
- 60Steps. 403 steps, 9 vague phrases
- 70Inputs and preconditions. Inputs and preconditions are listed
- 100Consistency. Name and required fields are in place
- 100Progress reporting. Reports progress
- medium Safety rules and hard prohibitions inside a skill: they belong in the system prompt, here they protect nothing
- low 27 top-level sections: this looks like several domains in one skill
- low The response is described with custom markup (5 tags): a typed call is more reliable
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
- +1No license
- +2Single-language instructions
- +3Description length 203: enough signal without eating the budget
- +4Structure: 75 headings
- +3Step-by-step instructions: 403 items
- +4Has examples (34 code blocks)
Quality base 70; lint remarks subtract, signals add up to 100. Result: 69.