📦
Audit History
security-engineer - 4 audits
Version comparison
Capability and finding changes across audited versions, newest first.
Audit version 4 Latest
Jul 24, 2026, 12:04 AM
Most critical static alerts are false positives caused by defensive examples, public references, and Markdown syntax. Actionable scanning, sensitive-data inspection, destructive removal, and an unaudited hidden-path script remain confirmed risks. The broad post-exploitation workflow creates additional high dual-use risk despite its authorization reminder.
2
Files scanned
416
Lines analyzed
36
Review items
0
False positives ignored
Confirmed security concerns (5)
High
Network scanning tools
- `nmap -sV -sC -O <target>` — 端口扫描+服务指纹+OS检测
The skill provides a ready-to-run Nmap command for service, operating-system, or broad port discovery. It enables active remote reconnaissance and can affect unauthorized targets.
High
Network scanning tools
nmap -sV -sC -O -A -T4 <target> -oA scan_result
The skill provides a ready-to-run Nmap command for service, operating-system, or broad port discovery. It enables active remote reconnaissance and can affect unauthorized targets.
High
Network scanning tools
nmap -p- --min-rate=1000 <target> -oG all_ports.txt
The skill provides a ready-to-run Nmap command for service, operating-system, or broad port discovery. It enables active remote reconnaissance and can affect unauthorized targets.
High
Network scanning tools
nmap -sV -p <ports> <target>
The skill provides a ready-to-run Nmap command for service, operating-system, or broad port discovery. It enables active remote reconnaissance and can affect unauthorized targets.
High
Offensive Post-Exploitation Workflow
The Web testing workflow explicitly progresses through exploitation, privilege escalation, and lateral movement, while also recommending parameter and dictionary brute forcing.
The offensive progression and brute-force guidance are explicit in the workflow. A general authorization reminder exists, but no per-action approval or technical scope control is defined.
Capability review items (28)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.
High
Hidden file in home directory
python3 ~/.hermes/skills/software-engineering/security-engineer/scripts/generate_pdf.py
The guide instructs execution of a Python script from a hidden home-directory path, but that script is absent from the audited package. This can execute unverified code outside the package boundary.
High
Ruby/shell backtick execution
- `nmap -sV -sC -O <target>` — 端口扫描+服务指纹+OS检测
This is an actionable active-scanning or injection-testing command against a user-selected target. It can disrupt services or probe systems outside an authorized scope.
High
Ruby/shell backtick execution
- **SQLMap**:`sqlmap -u <url> --data=<data> --batch --level=3 --risk=2`
This is an actionable active-scanning or injection-testing command against a user-selected target. It can disrupt services or probe systems outside an authorized scope.
High
Ruby/shell backtick execution
- **OWASP ZAP**:`zap-cli quick-scan --self-contained <url>`
This is an actionable active-scanning or injection-testing command against a user-selected target. It can disrupt services or probe systems outside an authorized scope.
High
Ruby/shell backtick execution
- **Nikto**:`nikto -h <target> -ssl -Format html -o report.html`
This is an actionable active-scanning or injection-testing command against a user-selected target. It can disrupt services or probe systems outside an authorized scope.
High
Ruby/shell backtick execution
- **Nessus / OpenVAS**:`openvas-start && gvm-cli --gmp-username admin --gmp-password <pass> socket -
The command starts OpenVAS and passes a scanner password on the command line. It changes service state and can expose the supplied password through shell history or process inspection.
High
Ruby/shell backtick execution
- **Nuclei**:`nuclei -u <target> -severity critical,high -o results.txt`
This is an actionable active-scanning or injection-testing command against a user-selected target. It can disrupt services or probe systems outside an authorized scope.
High
Ruby/shell backtick execution
- **Wazuh**:`/var/ossec/bin/wazuh-control status` → 告警查看 `cat /var/ossec/logs/alerts/alerts.json`
This actionable forensic command reads security alerts, memory data, or strings associated with credentials and secrets. Its output can expose sensitive operational data.
High
Ruby/shell backtick execution
- **Volatility**:`volatility -f memory.dump --profile=<profile> pslist` / `netscan`
This actionable forensic command reads security alerts, memory data, or strings associated with credentials and secrets. Its output can expose sensitive operational data.
High
Ruby/shell backtick execution
- **Strings**:`strings <binary> | grep -i 'password\|secret\|key\|http'`
This actionable forensic command reads security alerts, memory data, or strings associated with credentials and secrets. Its output can expose sensitive operational data.
High
Ruby/shell backtick execution
- **ClamAV**:`clamscan -r <path> --remove`
The skill supplies an actionable recursive ClamAV command with --remove. Executing it can delete files without quarantine, review, or per-file confirmation.
Medium
Hidden file access
python3 ~/.hermes/skills/software-engineering/security-engineer/scripts/generate_pdf.py
The regeneration command accesses and executes a script under ~/.hermes rather than a package-relative audited file. The external hidden path creates an integrity and provenance risk.
Medium
Ruby/shell backtick execution
- **Semgrep**:`semgrep --config=auto --config=p/r2c-security-audit <path>`
The skill presents a complete external command for execution against local files, services, logs, or images. Execution can read data or change local tool state and requires explicit approval.
Medium
Ruby/shell backtick execution
- **SonarQube**:`sonar-scanner -Dsonar.projectKey=<key> -Dsonar.sources=.`
The skill presents a complete external command for execution against local files, services, logs, or images. Execution can read data or change local tool state and requires explicit approval.
Medium
Ruby/shell backtick execution
- **Trivy**:`trivy fs --scanners vuln,secret,misconfig <path>`
The skill presents a complete external command for execution against local files, services, logs, or images. Execution can read data or change local tool state and requires explicit approval.
Medium
Ruby/shell backtick execution
- **Bandit**(Python):`bandit -r <path> -f json -o report.json`
The skill presents a complete external command for execution against local files, services, logs, or images. Execution can read data or change local tool state and requires explicit approval.
Medium
Ruby/shell backtick execution
- **Safety**(Python依赖):`safety check -r requirements.txt`
The skill presents a complete external command for execution against local files, services, logs, or images. Execution can read data or change local tool state and requires explicit approval.
Medium
Ruby/shell backtick execution
- **Trivy**:`trivy image <image>` / `trivy fs --scanners vuln,secret,misconfig <path>`
The skill presents a complete external command for execution against local files, services, logs, or images. Execution can read data or change local tool state and requires explicit approval.
Medium
Ruby/shell backtick execution
- **ELK Stack**:`curl -XGET 'localhost:9200/_search'` 查询日志
The skill presents a complete external command for execution against local files, services, logs, or images. Execution can read data or change local tool state and requires explicit approval.
Medium
Ruby/shell backtick execution
- **Wazuh**:`/var/ossec/bin/wazuh-control status` 检查状态
The skill presents a complete external command for execution against local files, services, logs, or images. Execution can read data or change local tool state and requires explicit approval.
Medium
Ruby/shell backtick execution
- **Grep/awk/sed**:日志快速过滤 `grep -E 'Failed password|Invalid user' /var/log/auth.log`
The skill presents a complete external command for execution against local files, services, logs, or images. Execution can read data or change local tool state and requires explicit approval.
Medium
Ruby/shell backtick execution
- **YARA**:`yara -r <rules.yar> <path>` 恶意文件扫描
The skill presents a complete external command for execution against local files, services, logs, or images. Execution can read data or change local tool state and requires explicit approval.
Medium
Ruby/shell backtick execution
- **OpenSCAP**:`oscap xccdf eval --profile xccdf_org.ssgproject.content_profile_<profile> --results
The skill presents a complete external command for execution against local files, services, logs, or images. Execution can read data or change local tool state and requires explicit approval.
Medium
Ruby/shell backtick execution
- **Lynis**:`lynis audit system --quick`
The skill presents a complete external command for execution against local files, services, logs, or images. Execution can read data or change local tool state and requires explicit approval.
Medium
Ruby/shell backtick execution
- **Osquery**:`osqueryi "SELECT * FROM processes WHERE name LIKE '%malware%';"`
The skill presents a complete external command for execution against local files, services, logs, or images. Execution can read data or change local tool state and requires explicit approval.
Medium
Ruby/shell backtick execution
- **Auditd**:`ausearch -m avc -ts today` / `aureport --summary`
The skill presents a complete external command for execution against local files, services, logs, or images. Execution can read data or change local tool state and requires explicit approval.
Medium
Ruby/shell backtick execution
- **YARA**:`yara -r <rules.yar> <path>`
The skill presents a complete external command for execution against local files, services, logs, or images. Execution can read data or change local tool state and requires explicit approval.
Medium
Ruby/shell backtick execution
- **Lsof**:`lsof -i -P -n` 查看网络连接
The skill presents a complete external command for execution against local files, services, logs, or images. Execution can read data or change local tool state and requires explicit approval.
Risk Factors
📁 Filesystem access (4)
⚙️ External commands (50)
SKILL.md:35-37 SKILL.md:37-40 SKILL.md:40-41 SKILL.md:41-42 SKILL.md:42-43 SKILL.md:43-44 SKILL.md:44-45 SKILL.md:45-51 SKILL.md:51-53 SKILL.md:53-61 SKILL.md:61-62 SKILL.md:62-63 SKILL.md:63-69 SKILL.md:69-71 SKILL.md:71-74 SKILL.md:74-75 SKILL.md:75-76 SKILL.md:76-77 SKILL.md:77-78 SKILL.md:78-90 SKILL.md:90-92 SKILL.md:92-95 SKILL.md:95-96 SKILL.md:96 SKILL.md:97-102 SKILL.md:102-104 SKILL.md:104-107 SKILL.md:107-108 SKILL.md:108-109 SKILL.md:109-113 SKILL.md:113-117 SKILL.md:117-119 SKILL.md:119-122 SKILL.md:122-123 SKILL.md:123-129 SKILL.md:129-131 SKILL.md:131-135 SKILL.md:135 SKILL.md:136-137 SKILL.md:137 SKILL.md:144-146 SKILL.md:146-150 SKILL.md:150 SKILL.md:151-152 SKILL.md:152 SKILL.md:157-159 SKILL.md:159-162 SKILL.md:162 SKILL.md:164-165 SKILL.md:165-166
🌐 Network access (9)
Detected Patterns
Network scanning toolsNetwork scanning toolsNetwork scanning toolsNetwork scanning tools
Audited by: codex
Jul 15, 2026, 03:42 PM
Most detections are false positives caused by Markdown code spans, defensive payload examples, and links to recognized security references. Confirmed risks include an unaudited home-directory script and active Nmap commands; semantic review also found destructive deletion, process-argument credential exposure, and a broad offensive workflow. Authorization guidance is present, but high-impact actions still need enforceable scope and confirmation controls.
2
Files scanned
416
Lines analyzed
12
Review items
0
False positives ignored
Confirmed security concerns (7)
High
Network scanning tools
- `nmap -sV -sC -O <target>` — 端口扫描+服务指纹+OS检测
The Nmap command performs active service, script, and operating-system probes against a target. The authorization guidance reduces misuse intent, but an incorrect scope can still disrupt or scan third-party systems.
High
Network scanning tools
nmap -sV -sC -O -A -T4 <target> -oA scan_result
This ready-to-run Nmap command enables aggressive detection, scripts, and operating-system fingerprinting. It can create substantial target traffic and expose services if executed outside the authorized scope.
High
Network scanning tools
nmap -p- --min-rate=1000 <target> -oG all_ports.txt
The command scans every TCP port at a minimum rate of 1,000 packets per second. This is an intrusive operation that can affect production systems or trigger defensive controls.
High
Network scanning tools
nmap -sV -p <ports> <target>
This command actively probes selected ports for service versions. It is legitimate within written scope, but remains a real network action with operational and authorization risk.
High
Broad Offensive Intrusion Workflow
The workflow directs agents through exploitation, privilege escalation, lateral movement, and password attacks. Prose authorization checks do not enforce targets or phase-specific approval.
The offensive stages and Burp Intruder password attack guidance are explicit. Authorization is required elsewhere, but no enforceable scope control is defined.
High
Destructive Malware Scan Option
The ClamAV example uses --remove during recursive scanning. False positives or an incorrect path could delete important files without quarantine or recovery.
The documented command explicitly combines recursive scanning with automatic removal. The line provides no confirmation, quarantine, backup, or recovery step.
Medium
Credential Exposure in Process Arguments
The OpenVAS example supplies a password through a command-line flag. Real credentials can remain visible in process listings, terminal logs, or shell history.
The --gmp-password flag is explicit, although the value is a placeholder. Replacing it with a real secret creates a well-known local exposure path.
Capability review items (2)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.
High
Hidden file in home directory
python3 ~/.hermes/skills/software-engineering/security-engineer/scripts/generate_pdf.py
The guide instructs Python to execute a generator under ~/.hermes, but the audited file tree contains no such script. A planted or stale file at that unaudited path could run with the agent's privileges.
High
Hidden file access
python3 ~/.hermes/skills/software-engineering/security-engineer/scripts/generate_pdf.py
This accesses a hidden home-directory path outside the audited files and immediately passes it to Python. The package provides no file whose contents can be verified at that path.
Risk Factors
📁 Filesystem access (4)
⚙️ External commands (50)
SKILL.md:35-37 SKILL.md:37-40 SKILL.md:40-41 SKILL.md:41-42 SKILL.md:42-43 SKILL.md:43-44 SKILL.md:44-45 SKILL.md:45-51 SKILL.md:51-53 SKILL.md:53-61 SKILL.md:61-62 SKILL.md:62-63 SKILL.md:63-69 SKILL.md:69-71 SKILL.md:71-74 SKILL.md:74-75 SKILL.md:75-76 SKILL.md:76-77 SKILL.md:77-78 SKILL.md:78-90 SKILL.md:90-92 SKILL.md:92-95 SKILL.md:95-96 SKILL.md:96 SKILL.md:97-102 SKILL.md:102-104 SKILL.md:104-107 SKILL.md:107-108 SKILL.md:108-109 SKILL.md:109-113 SKILL.md:113-117 SKILL.md:117-119 SKILL.md:119-122 SKILL.md:122-123 SKILL.md:123-129 SKILL.md:129-131 SKILL.md:131-135 SKILL.md:135 SKILL.md:136-137 SKILL.md:137 SKILL.md:144-146 SKILL.md:146-150 SKILL.md:150 SKILL.md:151-152 SKILL.md:152 SKILL.md:157-159 SKILL.md:159-162 SKILL.md:162 SKILL.md:164-165 SKILL.md:165-166
🌐 Network access (9)
Detected Patterns
Network scanning toolsNetwork scanning toolsNetwork scanning toolsNetwork scanning tools
Audited by: codex
Jul 15, 2026, 03:42 PM
Most detections are false positives caused by Markdown code spans, defensive payload examples, and links to recognized security references. Confirmed risks include an unaudited home-directory script and active Nmap commands; semantic review also found destructive deletion, process-argument credential exposure, and a broad offensive workflow. Authorization guidance is present, but high-impact actions still need enforceable scope and confirmation controls.
2
Files scanned
416
Lines analyzed
12
Review items
0
False positives ignored
Confirmed security concerns (7)
High
Network scanning tools
- `nmap -sV -sC -O <target>` — 端口扫描+服务指纹+OS检测
The Nmap command performs active service, script, and operating-system probes against a target. The authorization guidance reduces misuse intent, but an incorrect scope can still disrupt or scan third-party systems.
High
Network scanning tools
nmap -sV -sC -O -A -T4 <target> -oA scan_result
This ready-to-run Nmap command enables aggressive detection, scripts, and operating-system fingerprinting. It can create substantial target traffic and expose services if executed outside the authorized scope.
High
Network scanning tools
nmap -p- --min-rate=1000 <target> -oG all_ports.txt
The command scans every TCP port at a minimum rate of 1,000 packets per second. This is an intrusive operation that can affect production systems or trigger defensive controls.
High
Network scanning tools
nmap -sV -p <ports> <target>
This command actively probes selected ports for service versions. It is legitimate within written scope, but remains a real network action with operational and authorization risk.
High
Broad Offensive Intrusion Workflow
The workflow directs agents through exploitation, privilege escalation, lateral movement, and password attacks. Prose authorization checks do not enforce targets or phase-specific approval.
The offensive stages and Burp Intruder password attack guidance are explicit. Authorization is required elsewhere, but no enforceable scope control is defined.
High
Destructive Malware Scan Option
The ClamAV example uses --remove during recursive scanning. False positives or an incorrect path could delete important files without quarantine or recovery.
The documented command explicitly combines recursive scanning with automatic removal. The line provides no confirmation, quarantine, backup, or recovery step.
Medium
Credential Exposure in Process Arguments
The OpenVAS example supplies a password through a command-line flag. Real credentials can remain visible in process listings, terminal logs, or shell history.
The --gmp-password flag is explicit, although the value is a placeholder. Replacing it with a real secret creates a well-known local exposure path.
Capability review items (2)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.
High
Hidden file in home directory
python3 ~/.hermes/skills/software-engineering/security-engineer/scripts/generate_pdf.py
The guide instructs Python to execute a generator under ~/.hermes, but the audited file tree contains no such script. A planted or stale file at that unaudited path could run with the agent's privileges.
High
Hidden file access
python3 ~/.hermes/skills/software-engineering/security-engineer/scripts/generate_pdf.py
This accesses a hidden home-directory path outside the audited files and immediately passes it to Python. The package provides no file whose contents can be verified at that path.
Risk Factors
📁 Filesystem access (4)
⚙️ External commands (50)
SKILL.md:35-37 SKILL.md:37-40 SKILL.md:40-41 SKILL.md:41-42 SKILL.md:42-43 SKILL.md:43-44 SKILL.md:44-45 SKILL.md:45-51 SKILL.md:51-53 SKILL.md:53-61 SKILL.md:61-62 SKILL.md:62-63 SKILL.md:63-69 SKILL.md:69-71 SKILL.md:71-74 SKILL.md:74-75 SKILL.md:75-76 SKILL.md:76-77 SKILL.md:77-78 SKILL.md:78-90 SKILL.md:90-92 SKILL.md:92-95 SKILL.md:95-96 SKILL.md:96 SKILL.md:97-102 SKILL.md:102-104 SKILL.md:104-107 SKILL.md:107-108 SKILL.md:108-109 SKILL.md:109-113 SKILL.md:113-117 SKILL.md:117-119 SKILL.md:119-122 SKILL.md:122-123 SKILL.md:123-129 SKILL.md:129-131 SKILL.md:131-135 SKILL.md:135 SKILL.md:136-137 SKILL.md:137 SKILL.md:144-146 SKILL.md:146-150 SKILL.md:150 SKILL.md:151-152 SKILL.md:152 SKILL.md:157-159 SKILL.md:159-162 SKILL.md:162 SKILL.md:164-165 SKILL.md:165-166
🌐 Network access (9)
Detected Patterns
Network scanning toolsNetwork scanning toolsNetwork scanning toolsNetwork scanning tools
Audited by: codex
Jul 15, 2026, 03:42 PM
Most detections are false positives caused by Markdown code spans, defensive payload examples, and links to recognized security references. Confirmed risks include an unaudited home-directory script and active Nmap commands; semantic review also found destructive deletion, process-argument credential exposure, and a broad offensive workflow. Authorization guidance is present, but high-impact actions still need enforceable scope and confirmation controls.
2
Files scanned
416
Lines analyzed
12
Review items
0
False positives ignored
Confirmed security concerns (7)
High
Network scanning tools
- `nmap -sV -sC -O <target>` — 端口扫描+服务指纹+OS检测
The Nmap command performs active service, script, and operating-system probes against a target. The authorization guidance reduces misuse intent, but an incorrect scope can still disrupt or scan third-party systems.
High
Network scanning tools
nmap -sV -sC -O -A -T4 <target> -oA scan_result
This ready-to-run Nmap command enables aggressive detection, scripts, and operating-system fingerprinting. It can create substantial target traffic and expose services if executed outside the authorized scope.
High
Network scanning tools
nmap -p- --min-rate=1000 <target> -oG all_ports.txt
The command scans every TCP port at a minimum rate of 1,000 packets per second. This is an intrusive operation that can affect production systems or trigger defensive controls.
High
Network scanning tools
nmap -sV -p <ports> <target>
This command actively probes selected ports for service versions. It is legitimate within written scope, but remains a real network action with operational and authorization risk.
High
Broad Offensive Intrusion Workflow
The workflow directs agents through exploitation, privilege escalation, lateral movement, and password attacks. Prose authorization checks do not enforce targets or phase-specific approval.
The offensive stages and Burp Intruder password attack guidance are explicit. Authorization is required elsewhere, but no enforceable scope control is defined.
High
Destructive Malware Scan Option
The ClamAV example uses --remove during recursive scanning. False positives or an incorrect path could delete important files without quarantine or recovery.
The documented command explicitly combines recursive scanning with automatic removal. The line provides no confirmation, quarantine, backup, or recovery step.
Medium
Credential Exposure in Process Arguments
The OpenVAS example supplies a password through a command-line flag. Real credentials can remain visible in process listings, terminal logs, or shell history.
The --gmp-password flag is explicit, although the value is a placeholder. Replacing it with a real secret creates a well-known local exposure path.
Capability review items (2)
These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.
High
Hidden file in home directory
python3 ~/.hermes/skills/software-engineering/security-engineer/scripts/generate_pdf.py
The guide instructs Python to execute a generator under ~/.hermes, but the audited file tree contains no such script. A planted or stale file at that unaudited path could run with the agent's privileges.
High
Hidden file access
python3 ~/.hermes/skills/software-engineering/security-engineer/scripts/generate_pdf.py
This accesses a hidden home-directory path outside the audited files and immediately passes it to Python. The package provides no file whose contents can be verified at that path.
Risk Factors
📁 Filesystem access (4)
⚙️ External commands (50)
SKILL.md:35-37 SKILL.md:37-40 SKILL.md:40-41 SKILL.md:41-42 SKILL.md:42-43 SKILL.md:43-44 SKILL.md:44-45 SKILL.md:45-51 SKILL.md:51-53 SKILL.md:53-61 SKILL.md:61-62 SKILL.md:62-63 SKILL.md:63-69 SKILL.md:69-71 SKILL.md:71-74 SKILL.md:74-75 SKILL.md:75-76 SKILL.md:76-77 SKILL.md:77-78 SKILL.md:78-90 SKILL.md:90-92 SKILL.md:92-95 SKILL.md:95-96 SKILL.md:96 SKILL.md:97-102 SKILL.md:102-104 SKILL.md:104-107 SKILL.md:107-108 SKILL.md:108-109 SKILL.md:109-113 SKILL.md:113-117 SKILL.md:117-119 SKILL.md:119-122 SKILL.md:122-123 SKILL.md:123-129 SKILL.md:129-131 SKILL.md:131-135 SKILL.md:135 SKILL.md:136-137 SKILL.md:137 SKILL.md:144-146 SKILL.md:146-150 SKILL.md:150 SKILL.md:151-152 SKILL.md:152 SKILL.md:157-159 SKILL.md:159-162 SKILL.md:162 SKILL.md:164-165 SKILL.md:165-166
🌐 Network access (9)
Detected Patterns
Network scanning toolsNetwork scanning toolsNetwork scanning toolsNetwork scanning tools
Audited by: codex