📦
Audit History
broken-authentication-testing - 4 audits
Version comparison
Capability and finding changes across audited versions, newest first.
Audit version 4 Latest
Jul 6, 2026, 10:52 PM
Most external-command detections are false positives caused by Markdown code fences, not executable backtick usage. Several network and blocker findings are confirmed because the skill provides actionable guidance for credential attacks, rate-limit evasion, MFA bypass, session fixation, and password reset account takeover testing. No prompt injection attempt was found in SKILL.md.
1
Files scanned
477
Lines analyzed
26
Review items
0
False positives ignored
Confirmed security concerns (11)
High
System reconnaissance
"/login:username=^USER^&password=^PASS^:Invalid credentials"
The Hydra form string supports automated password attempts against a login endpoint. It is actionable brute-force guidance despite the authorized-testing framing.
High
System reconnaissance
- Previous valid OTP reuse
The OTP reuse item identifies a concrete MFA bypass condition. The use is dual-use and can support unauthorized access if misapplied.
High
System reconnaissance
# Try changing user parameter while using valid token
The instruction to change a user parameter while using a valid token targets password reset authorization logic. This can support account takeover.
High
Network reconnaissance
- Skip MFA step by direct URL access
The text recommends skipping the MFA step through direct URL access and response modification. This is actionable authentication bypass guidance.
High
Credential Stuffing And Evasion Guidance
The skill instructs testing with breached credential pairs and includes detection evasion steps such as slow rates, IP rotation, randomized user agents, and delays.
The guidance explicitly combines credential stuffing with evasion techniques. Written authorization is mentioned elsewhere, but the operational abuse content is clear.
High
MFA Bypass And OTP Attack Workflow
The skill describes OTP brute force, direct MFA step skipping, response modification, OTP reuse, and API version downgrade testing to bypass MFA controls.
The cited section provides multiple concrete MFA bypass techniques and validation steps. These techniques are dual-use but clearly actionable.
High
Session Fixation Account Takeover Scenario
The skill gives an attacker workflow where a victim logs in with a fixed attacker-controlled session and the attacker gains an authenticated session.
The section is framed as an attacker scenario and explains the victim interaction needed for session takeover. This is direct exploitation context.
High
Password Reset Account Takeover Testing
The skill includes reset-token user manipulation, host header injection, and admin email parameter tampering to test whether another account can be reset.
The cited steps target password reset authorization and email routing flaws that can lead to account takeover. The examples are specific and operational.
Medium
System reconnaissance
# Compare responses for valid vs invalid usernames
The text instructs comparing valid and invalid username responses. This is useful for authorized testing but can enable username enumeration.
Medium
System reconnaissance
# Invalid: "Invalid username" vs Valid: "Invalid password"
The text gives example differential errors for valid and invalid usernames. This can help enumerate accounts before credential attacks.
Medium
Network reconnaissance
# Recovery process
The recovery-process checks can expose ways to disable MFA through account recovery. Legitimate testing is possible, but the business-logic abuse risk is real.
Capability review items (13)
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
Hardcoded URL
https://target.com/login?SESSIONID=attacker_session
The URL appears inside an attacker session-fixation workflow involving a victim login. This is actionable account takeover guidance if used outside an authorized lab.
High
Hardcoded URL
https://target.com/reset?token=abc123&user=victim
The reset URL demonstrates changing a user parameter while using a valid reset token. That pattern can support password reset account takeover testing or abuse.
High
Hardcoded URL
https://target.com/reset?token=a1b2c3d4e5f6&email=admin@example.com
The reset URL changes the email parameter to an admin account and asks whether an admin password can be reset. This is direct account takeover guidance.
Medium
Hardcoded IP address
X-Forwarded-For: 127.0.0.1
The IP appears under rate-limiting bypass headers. Even though it is an example value, the surrounding guidance teaches evasion of IP-based controls.
Medium
Hardcoded IP address
X-Real-IP: 127.0.0.1
The IP appears under rate-limiting bypass headers. Even though it is an example value, the surrounding guidance teaches evasion of IP-based controls.
Medium
Hardcoded IP address
X-Originating-IP: 127.0.0.1
The IP appears under rate-limiting bypass headers. Even though it is an example value, the surrounding guidance teaches evasion of IP-based controls.
Medium
Hardcoded IP address
X-Client-IP: 127.0.0.1
The IP appears under rate-limiting bypass headers. Even though it is an example value, the surrounding guidance teaches evasion of IP-based controls.
Medium
Hardcoded IP address
X-Remote-IP: 127.0.0.1
The IP appears under rate-limiting bypass headers. Even though it is an example value, the surrounding guidance teaches evasion of IP-based controls.
Medium
Hardcoded IP address
True-Client-IP: 127.0.0.1
The IP appears under rate-limiting bypass headers. Even though it is an example value, the surrounding guidance teaches evasion of IP-based controls.
Medium
Hardcoded IP address
X-Forwarded-For: 192.168.1.1
The private IP is used in an account lockout bypass example that rotates X-Forwarded-For values. The context is evasion of authentication rate limits.
Medium
Hardcoded IP address
X-Forwarded-For: 192.168.1.2
The private IP is used in an account lockout bypass example that rotates X-Forwarded-For values. The context is evasion of authentication rate limits.
Low
Python HTTP libraries
response = requests.get("https://target.com/login")
The Python example performs repeated HTTP requests to a target login endpoint to collect session tokens. This is expected security-testing behavior but still creates network activity against a target application.
Low
Hardcoded URL
response = requests.get("https://target.com/login")
The hardcoded URL is a placeholder in a copyable token-collection example. It is not a hidden callback, but it normalizes active requests against a login endpoint.
Risk Factors
⚙️ External commands (40)
SKILL.md:47-60 SKILL.md:60-64 SKILL.md:64-70 SKILL.md:70-76 SKILL.md:76-81 SKILL.md:81-89 SKILL.md:89-93 SKILL.md:93-102 SKILL.md:102-108 SKILL.md:108-121 SKILL.md:121-125 SKILL.md:125-139 SKILL.md:139-145 SKILL.md:145-160 SKILL.md:160-166 SKILL.md:166-175 SKILL.md:175-179 SKILL.md:179-195 SKILL.md:195-201 SKILL.md:201-214 SKILL.md:214-218 SKILL.md:218-225 SKILL.md:225-231 SKILL.md:231-247 SKILL.md:247-253 SKILL.md:253-277 SKILL.md:277-281 SKILL.md:281-289 SKILL.md:289-295 SKILL.md:295-314 SKILL.md:314-333 SKILL.md:333-355 SKILL.md:355-370 SKILL.md:370-377 SKILL.md:377-405 SKILL.md:405-424 SKILL.md:424-430 SKILL.md:430-445 SKILL.md:445-451 SKILL.md:451-467
Detected Patterns
System reconnaissanceSystem reconnaissanceSystem reconnaissanceSystem reconnaissanceSystem reconnaissanceNetwork reconnaissanceNetwork reconnaissance
Audited by: codex
Jul 6, 2026, 10:52 PM
Most external-command detections are false positives caused by Markdown code fences, not executable backtick usage. Several network and blocker findings are confirmed because the skill provides actionable guidance for credential attacks, rate-limit evasion, MFA bypass, session fixation, and password reset account takeover testing. No prompt injection attempt was found in SKILL.md.
1
Files scanned
477
Lines analyzed
26
Review items
0
False positives ignored
Confirmed security concerns (11)
High
System reconnaissance
"/login:username=^USER^&password=^PASS^:Invalid credentials"
The Hydra form string supports automated password attempts against a login endpoint. It is actionable brute-force guidance despite the authorized-testing framing.
High
System reconnaissance
- Previous valid OTP reuse
The OTP reuse item identifies a concrete MFA bypass condition. The use is dual-use and can support unauthorized access if misapplied.
High
System reconnaissance
# Try changing user parameter while using valid token
The instruction to change a user parameter while using a valid token targets password reset authorization logic. This can support account takeover.
High
Network reconnaissance
- Skip MFA step by direct URL access
The text recommends skipping the MFA step through direct URL access and response modification. This is actionable authentication bypass guidance.
High
Credential Stuffing And Evasion Guidance
The skill instructs testing with breached credential pairs and includes detection evasion steps such as slow rates, IP rotation, randomized user agents, and delays.
The guidance explicitly combines credential stuffing with evasion techniques. Written authorization is mentioned elsewhere, but the operational abuse content is clear.
High
MFA Bypass And OTP Attack Workflow
The skill describes OTP brute force, direct MFA step skipping, response modification, OTP reuse, and API version downgrade testing to bypass MFA controls.
The cited section provides multiple concrete MFA bypass techniques and validation steps. These techniques are dual-use but clearly actionable.
High
Session Fixation Account Takeover Scenario
The skill gives an attacker workflow where a victim logs in with a fixed attacker-controlled session and the attacker gains an authenticated session.
The section is framed as an attacker scenario and explains the victim interaction needed for session takeover. This is direct exploitation context.
High
Password Reset Account Takeover Testing
The skill includes reset-token user manipulation, host header injection, and admin email parameter tampering to test whether another account can be reset.
The cited steps target password reset authorization and email routing flaws that can lead to account takeover. The examples are specific and operational.
Medium
System reconnaissance
# Compare responses for valid vs invalid usernames
The text instructs comparing valid and invalid username responses. This is useful for authorized testing but can enable username enumeration.
Medium
System reconnaissance
# Invalid: "Invalid username" vs Valid: "Invalid password"
The text gives example differential errors for valid and invalid usernames. This can help enumerate accounts before credential attacks.
Medium
Network reconnaissance
# Recovery process
The recovery-process checks can expose ways to disable MFA through account recovery. Legitimate testing is possible, but the business-logic abuse risk is real.
Capability review items (13)
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
Hardcoded URL
https://target.com/login?SESSIONID=attacker_session
The URL appears inside an attacker session-fixation workflow involving a victim login. This is actionable account takeover guidance if used outside an authorized lab.
High
Hardcoded URL
https://target.com/reset?token=abc123&user=victim
The reset URL demonstrates changing a user parameter while using a valid reset token. That pattern can support password reset account takeover testing or abuse.
High
Hardcoded URL
https://target.com/reset?token=a1b2c3d4e5f6&email=admin@example.com
The reset URL changes the email parameter to an admin account and asks whether an admin password can be reset. This is direct account takeover guidance.
Medium
Hardcoded IP address
X-Forwarded-For: 127.0.0.1
The IP appears under rate-limiting bypass headers. Even though it is an example value, the surrounding guidance teaches evasion of IP-based controls.
Medium
Hardcoded IP address
X-Real-IP: 127.0.0.1
The IP appears under rate-limiting bypass headers. Even though it is an example value, the surrounding guidance teaches evasion of IP-based controls.
Medium
Hardcoded IP address
X-Originating-IP: 127.0.0.1
The IP appears under rate-limiting bypass headers. Even though it is an example value, the surrounding guidance teaches evasion of IP-based controls.
Medium
Hardcoded IP address
X-Client-IP: 127.0.0.1
The IP appears under rate-limiting bypass headers. Even though it is an example value, the surrounding guidance teaches evasion of IP-based controls.
Medium
Hardcoded IP address
X-Remote-IP: 127.0.0.1
The IP appears under rate-limiting bypass headers. Even though it is an example value, the surrounding guidance teaches evasion of IP-based controls.
Medium
Hardcoded IP address
True-Client-IP: 127.0.0.1
The IP appears under rate-limiting bypass headers. Even though it is an example value, the surrounding guidance teaches evasion of IP-based controls.
Medium
Hardcoded IP address
X-Forwarded-For: 192.168.1.1
The private IP is used in an account lockout bypass example that rotates X-Forwarded-For values. The context is evasion of authentication rate limits.
Medium
Hardcoded IP address
X-Forwarded-For: 192.168.1.2
The private IP is used in an account lockout bypass example that rotates X-Forwarded-For values. The context is evasion of authentication rate limits.
Low
Python HTTP libraries
response = requests.get("https://target.com/login")
The Python example performs repeated HTTP requests to a target login endpoint to collect session tokens. This is expected security-testing behavior but still creates network activity against a target application.
Low
Hardcoded URL
response = requests.get("https://target.com/login")
The hardcoded URL is a placeholder in a copyable token-collection example. It is not a hidden callback, but it normalizes active requests against a login endpoint.
Risk Factors
⚙️ External commands (40)
SKILL.md:47-60 SKILL.md:60-64 SKILL.md:64-70 SKILL.md:70-76 SKILL.md:76-81 SKILL.md:81-89 SKILL.md:89-93 SKILL.md:93-102 SKILL.md:102-108 SKILL.md:108-121 SKILL.md:121-125 SKILL.md:125-139 SKILL.md:139-145 SKILL.md:145-160 SKILL.md:160-166 SKILL.md:166-175 SKILL.md:175-179 SKILL.md:179-195 SKILL.md:195-201 SKILL.md:201-214 SKILL.md:214-218 SKILL.md:218-225 SKILL.md:225-231 SKILL.md:231-247 SKILL.md:247-253 SKILL.md:253-277 SKILL.md:277-281 SKILL.md:281-289 SKILL.md:289-295 SKILL.md:295-314 SKILL.md:314-333 SKILL.md:333-355 SKILL.md:355-370 SKILL.md:370-377 SKILL.md:377-405 SKILL.md:405-424 SKILL.md:424-430 SKILL.md:430-445 SKILL.md:445-451 SKILL.md:451-467
Detected Patterns
System reconnaissanceSystem reconnaissanceSystem reconnaissanceSystem reconnaissanceSystem reconnaissanceNetwork reconnaissanceNetwork reconnaissance
Audited by: codex
Jun 30, 2026, 12:33 PM
The static backtick, URL, IP address, C2, and weak-crypto findings are mostly false positives caused by fenced Markdown examples and placeholder values. However, the skill gives operational guidance for brute force testing, credential stuffing, detection evasion, MFA bypass, JWT tampering, and reset-token abuse, so publication should require human review and stronger safety constraints.
1
Files scanned
477
Lines analyzed
7
Review items
2
False positives ignored
Capability review items (5)
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
Offensive Credential Attack Guidance
The skill includes concrete workflows for brute force testing and credential stuffing, including Hydra usage, breached credential pairs, response analysis, and successful login detection. This is dual-use penetration testing content, but it can materially enable account takeover if used without authorization.
The lines explicitly describe automated login attacks and credential stuffing workflows. The document states authorization requirements elsewhere, but the attack steps are still operational.
High
Authentication Bypass and Token Tampering Playbook
The skill documents MFA bypass approaches, API downgrade testing, OTP brute force, JWT none-algorithm tampering, and bearer token submission. These techniques can support valid assessments, but they also provide direct bypass methods against real accounts.
The cited sections contain explicit bypass and token modification instructions. The context is security testing, but the instructions are directly actionable.
High
Rate Limit Evasion Guidance
The skill advises IP rotation, user-agent randomization, delays, X-Forwarded-For bypass testing, and per-attempt IP changes. This guidance can help validate defenses, but it also teaches evasion of controls that protect login endpoints.
The content directly discusses bypassing lockouts and rate limits with source IP manipulation. This is high-risk even when presented as an authorized test.
Medium
Session Fixation and Password Reset Abuse Scenarios
The skill provides attack scenarios for fixed sessions, host header manipulation, reset-token user parameter changes, and resetting another user's password with a test token. These are valid test cases but expose sensitive abuse workflows.
The cited examples describe specific account takeover paths. The content remains framed as testing, so this is moderate to high dual-use risk rather than confirmed malicious intent.
Low
Placeholder Network Values Are Not Exfiltration
The hardcoded URLs and IP addresses use placeholder domains or local/private addresses for test examples. I found no evidence that the skill sends data to an author-controlled endpoint.
The URLs point to target.com examples and the IPs are loopback or private examples. There is no author-controlled collection endpoint in the reviewed file.
Static false positives ignored (2)
These static matches were dismissed by semantic review or matched schema-only tokens, so they are shown for transparency but do not drive the quality score.
Low
Static Backtick Command Findings Are Mostly False Positives
The external command detector flagged many fenced Markdown blocks as Ruby or shell backtick execution. The reviewed file is documentation, not executable Ruby code, so these locations do not show direct command execution by the skill.
The flagged backticks are Markdown code fences, not language-level execution syntax. The examples are still high-risk by content, but not because the skill executes commands.
Low
C2 and Weak-Crypto Static Blockers Are Contextual False Positives
The C2 keyword detections correspond to bearer tokens in JWT examples, and the weak-crypto detections appear in authentication terminology and JWT algorithm discussion. I found no evidence of malware control logic or implemented weak cryptography.
The suspicious tokens are part of explanatory web security examples. The file does not contain executable malware logic or cryptographic implementation code.
Risk Factors
⚙️ External commands (40)
SKILL.md:47-60 SKILL.md:60-64 SKILL.md:64-70 SKILL.md:70-76 SKILL.md:76-81 SKILL.md:81-89 SKILL.md:89-93 SKILL.md:93-102 SKILL.md:102-108 SKILL.md:108-121 SKILL.md:121-125 SKILL.md:125-139 SKILL.md:139-145 SKILL.md:145-160 SKILL.md:160-166 SKILL.md:166-175 SKILL.md:175-179 SKILL.md:179-195 SKILL.md:195-201 SKILL.md:201-214 SKILL.md:214-218 SKILL.md:218-225 SKILL.md:225-231 SKILL.md:231-247 SKILL.md:247-253 SKILL.md:253-277 SKILL.md:277-281 SKILL.md:281-289 SKILL.md:289-295 SKILL.md:295-314 SKILL.md:314-333 SKILL.md:333-355 SKILL.md:355-370 SKILL.md:370-377 SKILL.md:377-405 SKILL.md:405-424 SKILL.md:424-430 SKILL.md:430-445 SKILL.md:445-451 SKILL.md:451-467
Detected Patterns
Automated Login Attack ExamplesCredential Stuffing and Evasion StepsToken and MFA Bypass Techniques
Audited by: codex
Feb 25, 2026, 04:31 AM
This is a legitimate penetration testing and security assessment skill. All 76 static findings are false positives: the detected 'external_commands' and 'network' patterns are documentation code examples showing legitimate security testing techniques (hydra commands, HTTP request examples, JWT attack demonstrations). The skill properly includes legal requirements for authorization and scope limitations. No actual executable malware or exploit code is present.
1
Files scanned
477
Lines analyzed
2
Review items
0
False positives ignored
Risk Factors
⚙️ External commands (40)
SKILL.md:47-60 SKILL.md:60-64 SKILL.md:64-70 SKILL.md:70-76 SKILL.md:76-81 SKILL.md:81-89 SKILL.md:89-93 SKILL.md:93-102 SKILL.md:102-108 SKILL.md:108-121 SKILL.md:121-125 SKILL.md:125-139 SKILL.md:139-145 SKILL.md:145-160 SKILL.md:160-166 SKILL.md:166-175 SKILL.md:175-179 SKILL.md:179-195 SKILL.md:195-201 SKILL.md:201-214 SKILL.md:214-218 SKILL.md:218-225 SKILL.md:225-231 SKILL.md:231-247 SKILL.md:247-253 SKILL.md:253-277 SKILL.md:277-281 SKILL.md:281-289 SKILL.md:289-295 SKILL.md:295-314 SKILL.md:314-333 SKILL.md:333-355 SKILL.md:355-370 SKILL.md:370-377 SKILL.md:377-405 SKILL.md:405-424 SKILL.md:424-430 SKILL.md:430-445 SKILL.md:445-451 SKILL.md:451-467
Audited by: claude