Getting it into your agent
It runs from inside its repository, so the clone comes first — what it calls does not travel with the file alone.
git clone --depth 1 https://github.com/kalpmodi/akiranpx agentmods add skills/kalpmodi/akira/oauth-attacksWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/kalpmodi/akira/oauth-attacks)<a href="https://agentmods.dev/skills/kalpmodi/akira/oauth-attacks"><img src="https://agentmods.dev/badge/skills/kalpmodi/akira/oauth-attacks.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00096 | $0.04482 |
| Opus 5 | $0.00048 | $0.02241 |
| Sonnet 5 | $0.00019 | $0.00896 |
| Haiku 4.5 | $0.00010 | $0.00448 |
Grade C, and why
oauth-attacks scanned grade C with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Downloads and executes remote codehighSupply chain
curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.
curl https://<target>/.well-known/openid-configuration | python3 -m json.tool Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl https://<target>/.well-known/openid-configuration How it starts
The opening of the file, as written. The whole thing — 415 lines — stays where its author put it; the contents beside it link to each section on GitHub.
OAuth 2.0 / OIDC Attack Playbook
Philosophy
OAuth bugs are the highest-paying web vulns on HackerOne. They chain directly to account takeover. The attacker controls the redirect. Find one open redirect on the OAuth domain and you own the flow. Never claim "account takeover" without demonstrating actual token receipt for another user's account.
Arguments
<target> - domain with OAuth (e.g. app.target.com)
<focus> - optional: REDIRECT / CSRF / TOKEN-LEAK / JWT / PKCE / FULL
Phase 0: Smart Intake
source ~/.claude/skills/_shared/phase0.sh
source ~/.claude/skills/_shared/signals.sh
p0_init_vars "$1"
p0_state_gate "HARVEST" || exit 0
p0_read_relay recon secrets
p0_read_memory
p0_read_hypotheses
# Narrow endpoints + wayback to OAuth-relevant paths
INTERESTING_ENDPOINTS=$(echo "$INTERESTING_ENDPOINTS" | grep -i "oauth\|auth\|login\|sso\|connect\|token\|authorize" | head -20)
WAYBACK_ENDPOINTS=$(echo "$WAYBACK_ENDPOINTS" | grep -i "oauth\|auth\|token" | head -10)
JWT_COUNT=$(echo "$JWT_TOKENS" | grep -c "ey" 2>/dev/null || echo 0)
TECH_STACK=$KNOWN_TECH
echo "=== PHASE 0 OAUTH INTAKE: $TARGET ==="
echo "State: $STATE | JWT tokens already found: $JWT_COUNT"
echo "OAuth-related endpoints from recon: $(echo "$INTERESTING_ENDPOINTS" | wc -l)"
echo "Tech stack: $TECH_STACK"
echo "Top hypothesis: $TOP_HYPO_LABEL [$TOP_HYPO_PROB%]"
echo "ATW flagged (avoid): ${ATW_FLAGGED:-none}"
Execution Manifest
MANIFEST=$(cat << 'MANIFEST_EOF'
{
"phase": "oauth-attacks",
"generated_at": "YYYY-MM-DD HH:MM",
"items": [
{"id":"o01","tool":"oidc-discovery","target":"<target>/.well-known/openid-configuration","reason":"enumerate all OAuth endpoints + supported flows","priority":"MUST","status":"pending","skip_reason":null},
{"id":"o02","tool":"redirect-uri-bypass","target":"<oauth-authorize-endpoint>","reason":"code interception via loose redirect_uri validation","priority":"MUST","status":"pending","skip_reason":null},
{"id":"o03","tool":"state-csrf-test","target":"<oauth-callback-endpoint>","reason":"missing/bypassable state = CSRF account linking","priority":"MUST","status":"pending","skip_reason":null},
{"id":"o04","tool":"pkce-downgrade","target":"<oauth-authorize-endpoint>","reason":"PKCE not enforced = code interception without verifier","priority":"MUST","status":"pending","skip_reason":null},
{"id":"o05","tool":"jwt-claim-attack","target":"<access-token>","reason":"alg:none / RS256->HS256 confusion on OAuth tokens","priority":"MUST","status":"pending","skip_reason":"set to skipped if no JWT tokens found"},
{"id":"o06","tool":"token-leakage-referer","target":"<oauth-callback-page>","reason":"auth code in Referer to third-party analytics","priority":"SHOULD","status":"pending","skip_reason":null},
{"id":"o07","tool":"open-redirect-chain","target":"<target> redirectors","reason":"find open redirect + chain with redirect_uri bypass","priority":"MUST","status":"pending","skip_reason":null},
{"id":"o08","tool":"scope-escalation","target":"<token-endpoint>","reason":"request read scope, check if write/admin included","priority":"SHOULD","status":"pending","skip_reason":null},
{"id":"o09","tool":"jwks-uri-hijack","target":"<jku-claim-in-token>","reason":"attacker-controlled JWKS = forge any token","priority":"IF_TIME","status":"pending","skip_reason":"skip if no jku claim in token header"}
]
}
MANIFEST_EOF
)
jq --argjson m "$MANIFEST" '.scalpel.active_manifest = $m' $SESSION > /tmp/s.json && mv /tmp/s.json $SESSION
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 8d ago First seen · 415 lines · 96 tokens per session scan C 2744049eefda
oauth-attacks is a skill published in the GitHub repository kalpmodi/akira (21 stars, last pushed 1mo ago), licensed MIT. It adds 96 tokens to every session and 4,482 once invoked, about $0.0005 per session on Opus 5. A static security scan graded it C with 2 findings (downloads and executes remote code, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
hunt-llm-ai
Hunt LLM/AI feature bugs — prompt injection, indirect injection, exfiltration viatool-use/markdown, ASCII smuggling, agentic AI security (OWASP Agentic Apps 2026, ASI01-ASI10). Patterns: direct injection ('ignore previous instructions'), indirect injection via documents/web pages/email the model reads, ASCII smuggling…
performing-aws-privilege-escalation-assessment
Performing authorized privilege escalation assessments in AWS environments to identify IAM misconfigurations that allow users or roles to elevate their permissions using Pacu, CloudFox, Principal Mapper, and manual IAM policy analysis techniques.
performing-cloud-penetration-testing-with-pacu
Performing authorized AWS penetration testing using Pacu, the open-source AWS exploitation framework, to enumerate IAM configurations, discover privilege escalation paths, test credential harvesting, and validate security controls through systematic attack simulation.
conducting-cloud-penetration-testing
This skill outlines methodologies for performing authorized penetration testing against AWS, Azure, and GCP cloud environments. It covers understanding the shared responsibility model for testing scope, leveraging cloud-specific attack tools like Pacu and ScoutSuite, exploiting IAM misconfigurations, testing for SSRF…
performing-kubernetes-penetration-testing
Kubernetes penetration testing systematically evaluates cluster security by simulating attacker techniques against the API server, kubelet, etcd, pods, RBAC, network policies, and secrets. Using tools.
conducting-full-scope-red-team-engagement
Plan and execute a comprehensive red team engagement covering reconnaissance through post-exploitation using MITRE ATT&CK-aligned TTPs to evaluate an organization's detection and response capabilities.