Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Dupcheck · h-mmer | DecimalAI

h-mmer/dupcheck

SkillScore45Maintained

Check if a vulnerability has already been reported. Searches platform hacktivity + local findings. Usage: /dupcheck <vuln_type> e.g. /dupcheck XSS in search endpoint

Securityby h-mmer

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

Security & AppSec pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/h-mmer-dupcheck/SKILL.md

decimalai skills pull h-mmer-dupcheck --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/h-mmer-dupcheck/SKILL.md https://app.decimal.ai/s/h-mmer-dupcheck/SKILL.md

Raw file: /s/h-mmer-dupcheck/SKILL.md ↗ · machine-readable summary at /s/h-mmer-dupcheck.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 56d ago

—+2% tokens0% turns45% of whole cases pass with skill (10/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-08	✗→✓	▲ Improved	-19%	0%
case-10	✗→✓	▲ Improved	-31%	0%
case-11	✗→✓	▲ Improved	-11%	0%
case-12	✗→✓	▲ Improved	-19%	0%
case-18	✗→✓	▲ Improved	-7%	0%

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/1/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

H-mmer / pentest-agents

providers/codex/.agents/skills/dupcheck/SKILL.md

View on GitHub ↗

Imported 8/1/2026Version v1

Skill contents

SKILL.md

Check for duplicate reports: $ARGUMENTS

Determine the platform and program from scope.yaml in the current directory.
Use bounty-platforms MCP tool search_hacktivity with platform, program, and "$ARGUMENTS" as the query.
Also search local findings: uv run python3 ../../tools/dedup_findings.py --stats --db findings.json
Read hacktivity.md if it exists and grep for related terms.
Report:

Exact or near matches from hacktivity (potential duplicates)
Related reports that might overlap
If the area appears heavily reported (high duplicate risk)
Verdict: likely unique, possible duplicate, or high duplicate risk

Writeup Cross-Reference (if writeup-search MCP is available)

After checking local findings, also search the writeup database:

Use search_writeups MCP tool with "<finding description> <target>"
If similar writeups exist, assess whether your finding is novel or a known pattern
Mention relevant prior art in the "Known Techniques" section of the report

Top-Tier Duplicate Analysis

Duplicate risk is about overlap of exploit primitive and affected asset, not keyword similarity.

Report four verdict fields:

same_asset_same_primitive: likely duplicate unless your impact is strictly stronger
same_primitive_different_asset: possible duplicate; explain scope difference and novelty
same_asset_different_primitive: usually unique; prove a different root cause
known_class_new_chain: often worth reporting if the chain reaches a new impact tier

Check disclosed writeups for patch language and response tone. If triagers historically close this class as N/A, require chain proof before submission. If public reports stop at a weaker impact, frame your report around the new capability, not the shared first step.

Illustrative Want to see what h-mmer/dupcheck would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Latest benchmark · publisher runVerified

8/3/2026, 07:17 UTC · 22 cases

Measured on 14 comparable cases — below the 22 needed to publish a lift, so this result is shown here but not on the registry card or leaderboard.

Lift

+18%

45% vs 27%

Agent turns

1.0 vs 1.0

Tokens

+2%

1,776.07 vs 1,745.86

fail → pass · 6pass → pass · 4fail → fail · 10pass → fail · 2

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-05 We have a local SQLite or JSON database of past findings at `findings.json`. Run our dedicated Python deduplication script against this file to gather statistics on previous reports before filing a new SQLi.	fail→fail	10,191	5,592	-45%	1	1	0%	2,147	709	-67%	0	0	—
case-06 I want to cross-reference a novel HTTP Request Smuggling finding against disclosed community writeups to see if this pattern is known or novel. Run a writeup search and indicate prior art.	fail→fail	11,354	5,350	-53%	1	1	0%	1,764	782	-56%	0	0	—
case-07 We maintain a local markdown file named `hacktivity.md` in our workspace. Search this file for terms related to OAuth account takeover to evaluate past program history.	pass→fail	22,554	17,842	-21%	1	1	0%	3,746	793	-79%	0	0	—
case-08 We found an XSS on `app.example.com/search` using `<script>alert(1)</script>`. Another report already exists on `app.example.com/search` using `javascript:alert(1)`. Provide the explicit duplicate risk classification for having identical asset and identical primitive.	fail→pass	7,946	4,039	-49%	1	1	0%	1,409	1,142	-19%	0	0	—
case-01 I just found an IDOR on the profile update endpoint (`POST /api/v1/user/profile`). Before I submit this report to the program, can you check if this issue or similar ones have already been reported? Search through public program disclosures, writeups, and our local findings to give me an overall assessment of duplicate risk and a verdict on uniqueness.	fail→fail	13,675	5,218	-62%	1	1	0%	2,149	751	-65%	0	0	—
case-02 Please run a duplicate check for an unauthenticated SSRF vulnerability in the webhook endpoint of our target app. Look up existing hacktivity, local records, and public technical writeups for this program to see if this exploit area is heavily reported and summarize whether my finding is likely unique or overlapping.	fail→fail	17,002	31,864	+87%	1	1	0%	1,700	1,963	+15%	0	0	—
case-03 I'm preparing a bug report for a host header injection issue leading to password reset poisoning on `auth.example.com`. Can you check if this target asset and primitive match any previous reports or known writeups? I need an evaluation of any prior art and a clear risk assessment before filing.	fail→fail	17,672	21,536	+22%	1	1	0%	1,696	2,937	+73%	0	0	—
case-04 I discovered an open redirect on `https://sub.target.org/login?goto=https://evil.com`. Before drafting the submission, determine the program scope details from the local YAML config and check public disclosures for matching reports.	fail→fail	14,187	11,075	-22%	1	1	0%	1,707	1,914	+12%	0	0	—
case-09 We found a CSRF on `billing.example.com/update` leading to account takeover. There is a known disclosure of the exact same CSRF pattern on `settings.example.com/profile`. Assess the risk classification when the primitive matches but the target asset differs.	fail→fail	13,596	19,484	+43%	1	1	0%	2,267	636	-72%	0	0	—
case-10 On `api.example.com/v1/user`, there is a disclosed IDOR. We found a GraphQL field suggestion authorization bypass on `api.example.com/v1/user`. How should this be classified when the asset matches but the root cause primitive is entirely distinct?	fail→pass	11,554	6,156	-47%	1	1	0%	1,958	1,346	-31%	0	0	—
case-11 A known low-severity information disclosure on `admin.example.com` was previously closed. We chained that disclosure with an unauthenticated SSRF to achieve Remote Code Execution. What is the duplicate verdict classification for extending a known class into a higher impact chain?	fail→pass	11,692	6,747	-42%	1	1	0%	1,853	1,640	-11%	0	0	—
case-12 When reviewing disclosed reports for a target program, we noticed triagers historically close CORS misconfiguration issues as Informative or N/A. What requirement must be met before submitting a new CORS finding?	fail→pass	7,868	3,911	-50%	1	1	0%	1,369	1,109	-19%	0	0	—
case-13 Public writeups for `target.com` show that reflected XSS was previously reported but stopped at showing `alert(1)`. We escalation this XSS to exfiltrate session tokens and account takeover. How should the submission be framed?	pass→pass	15,473	24,313	+57%	1	1	0%	1,801	2,463	+37%	0	0	—
case-14 When evaluating duplicate risk for a rate limiting bypass on the authentication endpoint, deliver the full four-field risk analysis structure so triagers can assess overlap.	fail→fail	13,356	8,194	-39%	1	1	0%	2,291	683	-70%	0	0	—
case-15 Perform a platform search across disclosed program hacktivity for 'GraphQL Introspection' on the current project program.	fail→fail	10,371	12,495	+20%	1	1	0%	1,736	638	-63%	0	0	—
case-16 When writeups exist for a similar cache poisoning attack, in which specific section of the final report should the relevant prior art be referenced?	fail→fail	10,007	3,384	-66%	1	1	0%	1,636	663	-59%	0	0	—
case-17 Summarize duplicate risk for an arbitrary file read in the export feature after scanning hacktivity and local databases. What are the four summary risk levels to conclude with?	pass→fail	10,773	5,882	-45%	1	1	0%	1,709	852	-50%	0	0	—
case-18 Should duplicate risk be evaluated based on shared keywords in report titles or based on technical characteristics? Explain the primary criteria.	fail→pass	12,512	8,096	-35%	1	1	0%	1,813	1,686	-7%	0	0	—
case-19 What exact flags should be passed when invoking the local deduplication python script against `findings.json`?	fail→pass	5,441	2,897	-47%	1	1	0%	946	767	-19%	0	0	—
case-20 Calculate the CVSS v3.1 base score for a reflected XSS vulnerability found in the `q` parameter on `search.example.com` where user interaction is required.	pass→pass	9,130	9,502	+4%	1	1	0%	1,912	2,318	+21%	0	0	—
case-21 Provide secure coding remediation guidance in Python Django to fix an SQL injection vulnerability in a user filter query.	pass→pass	14,777	15,113	+2%	1	1	0%	2,811	3,322	+18%	0	0	—
case-22 Format a bug bounty report markdown template with standard sections like Description, Impact, Steps to Reproduce, and Mitigation for submission to HackerOne.	pass→pass	10,366	6,894	-33%	1	1	0%	1,831	1,595	-13%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 14 counted toward the lift figure. The other 8 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +18 percentage points is the difference between those two pass rates over the 14 comparable cases. 3 cases got worse with the skill loaded, and they are included in that figure.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 27.3%
Pass rate with the skill: 45.5%
Cases attempted: 22
Cases counted toward lift: 14
Cases that regressed: 3
Skill version measured: v1
Run completed: 8/3/2026, 07:17 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-08

Prompt

We found an XSS on `app.example.com/search` using `<script>alert(1)</script>`. Another report already exists on `app.example.com/search` using `javascript:alert(1)`. Provide the explicit duplicate risk classification for having identical asset and identical primitive.

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run45

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns+16% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 55 days ago

Other measured skills in the registry, with their headline benchmark lift.