Skip to content

borghei/Claude-Skills / engineering/security-audit

security-audit

Unverified

Multi-agent source security audit with a coverage ledger, independently verified findings, and machine-checked records. Use when asked to audit a codebase, find vulnerabilities, or run a source-level pen test.

assets/sample_coverage_ledger.json

View on GitHub
{
  "schema_version": 1,
  "run_id": "orders-api-run-1",
  "units": [
    {
      "coverage_id": "cu-0fd9027294b4ff3c369d",
      "refs": {
        "surface": "repository",
        "boundary": "repository#baseline",
        "subsystem": "repository",
        "attack_class": "attack-classes.md#Baseline exposures"
      },
      "label": "Whole repository: baseline exposures",
      "starting_paths": [
        "config/settings.py",
        "src/app.py"
      ],
      "class_blocks": [
        "attack-classes.md#Baseline exposures"
      ],
      "excluded_blocks": [],
      "origin": "none",
      "wave": 1,
      "status": "covered",
      "owner": "hunter-04",
      "reviewed_paths": [
        ".gitignore",
        "config/settings.py",
        "requirements.lock",
        "src/app.py"
      ],
      "checks": [
        {
          "agent_id": "hunter-04",
          "reviewed_paths": [
            "config/settings.py",
            ".gitignore",
            "requirements.lock",
            "src/app.py"
          ],
          "invariant": "No shipped secret, open debug switch, or unauthenticated diagnostic route",
          "method": "source",
          "result": "every item checked; debug is gated on an environment value read once at start; no secrets or unprotected routes found",
          "evidence_file": null
        }
      ],
      "fingerprints": [],
      "unresolved": [],
      "attempts": [],
      "hardening": [
        "Pin the two dependencies in requirements.lock that use a version range"
      ]
    },
    {
      "coverage_id": "cu-4b2c324566475f50b679",
      "refs": {
        "surface": "src/api/orders.py#GET /orders/search",
        "boundary": "src/db/orders.py#search_orders",
        "subsystem": "services/orders",
        "attack_class": "attack-classes.md#Injection"
      },
      "label": "Order search: query construction from the q and sort parameters",
      "starting_paths": [
        "src/api/orders.py",
        "src/db/orders.py"
      ],
      "class_blocks": [
        "attack-classes.md#Injection"
      ],
      "excluded_blocks": [],
      "origin": "none",
      "wave": 1,
      "status": "candidate",
      "owner": "hunter-03",
      "reviewed_paths": [
        "src/api/orders.py",
        "src/api/schemas.py",
        "src/db/orders.py"
      ],
      "checks": [
        {
          "agent_id": "hunter-03",
          "reviewed_paths": [
            "src/api/orders.py",
            "src/db/orders.py"
          ],
          "invariant": "Caller text never becomes SQL structure",
          "method": "source",
          "result": "q is bound as a parameter; sort is interpolated into ORDER BY and looked suspicious",
          "evidence_file": null
        },
        {
          "agent_id": "verifier-03",
          "reviewed_paths": [
            "src/db/orders.py",
            "src/api/schemas.py"
          ],
          "invariant": "The sort value is constrained before interpolation",
          "method": "source",
          "result": "schemas.py restricts sort to a four-value enum validated before the handler runs; no caller text reaches the query",
          "evidence_file": null
        }
      ],
      "fingerprints": [
        "orders:search:sort-interpolation"
      ],
      "unresolved": [],
      "attempts": []
    },
    {
      "coverage_id": "cu-4cd95d7167c66d6d8a54",
      "refs": {
        "surface": "src/api/orders.py#POST /orders/{id}/refund",
        "boundary": "src/services/refunds.py#apply_refund",
        "subsystem": "services/orders",
        "attack_class": "attack-classes.md#Business logic"
      },
      "label": "Refund flow: amount limits and repeated refunds",
      "starting_paths": [
        "src/services/refunds.py"
      ],
      "class_blocks": [
        "attack-classes.md#Business logic"
      ],
      "excluded_blocks": [],
      "origin": "critic",
      "wave": 2,
      "status": "covered",
      "owner": "hunter-06",
      "reviewed_paths": [
        "src/db/orders.py",
        "src/services/refunds.py",
        "tests/services/test_refunds.py"
      ],
      "checks": [
        {
          "agent_id": "hunter-06",
          "reviewed_paths": [
            "src/services/refunds.py",
            "src/db/orders.py",
            "tests/services/test_refunds.py"
          ],
          "invariant": "Total refunded never exceeds the amount paid, including under concurrent requests",
          "method": "source",
          "result": "refund and balance update run in one transaction with a row lock; negative and over-limit amounts are rejected before the write",
          "evidence_file": null
        }
      ],
      "fingerprints": [],
      "unresolved": [],
      "attempts": [
        {
          "wave": 1,
          "status": "covered",
          "owner": "hunter-05",
          "reviewed_paths": [
            "src/services/refunds.py"
          ],
          "checks": [
            {
              "agent_id": "hunter-05",
              "reviewed_paths": [
                "src/services/refunds.py"
              ],
              "invariant": "Refund amount is validated",
              "method": "source",
              "result": "amount is checked against the order total",
              "evidence_file": null
            }
          ],
          "fingerprints": [],
          "unresolved": [],
          "reopen_reason": "closed without reading the concurrent path in src/db/orders.py that the balance update uses"
        }
      ]
    },
    {
      "coverage_id": "cu-a8cbdb7fe84db79cd728",
      "refs": {
        "surface": "src/api/reports.py#GET /reports/export",
        "boundary": "src/authz/policy.py#require_role",
        "subsystem": "services/reporting",
        "attack_class": "attack-classes.md#Feature abuse and data exposure"
      },
      "label": "Report export: row scope of exported data",
      "starting_paths": [
        "src/api/reports.py"
      ],
      "class_blocks": [
        "attack-classes.md#Feature abuse and data exposure"
      ],
      "excluded_blocks": [],
      "origin": "none",
      "wave": 1,
      "status": "deferred",
      "owner": null,
      "reviewed_paths": [],
      "checks": [],
      "fingerprints": [],
      "unresolved": [
        "budget_cannot_fund_hunters"
      ],
      "attempts": []
    },
    {
      "coverage_id": "cu-b0309fd3c832ac96155e",
      "refs": {
        "surface": "src/api/auth.py#POST /password/reset",
        "boundary": "src/mail/links.py#build_reset_url",
        "subsystem": "services/identity",
        "attack_class": "web-protocol-and-auth.md#Host and forwarded-header trust"
      },
      "label": "Password reset: origin of the host used in emailed links",
      "starting_paths": [
        "src/api/auth.py",
        "src/mail/links.py"
      ],
      "class_blocks": [
        "web-protocol-and-auth.md#Host and forwarded-header trust",
        "web-protocol-and-auth.md#Password reset and account recovery"
      ],
      "excluded_blocks": [],
      "origin": "none",
      "wave": 1,
      "status": "candidate",
      "owner": "hunter-02",
      "reviewed_paths": [
        "config/settings.py",
        "deploy/ingress.yaml",
        "src/api/auth.py",
        "src/mail/links.py"
      ],
      "checks": [
        {
          "agent_id": "hunter-02",
          "reviewed_paths": [
            "src/api/auth.py",
            "src/mail/links.py",
            "config/settings.py"
          ],
          "invariant": "Reset links are built from a configured origin, never from request headers",
          "method": "source",
          "result": "build_reset_url uses request.host when PUBLIC_BASE_URL is unset; settings.py leaves it unset by default",
          "evidence_file": null
        },
        {
          "agent_id": "verifier-02",
          "reviewed_paths": [
            "src/mail/links.py",
            "deploy/ingress.yaml"
          ],
          "invariant": "The ingress in the repository fixes the Host header before it reaches the app",
          "method": "source",
          "result": "deploy/ingress.yaml routes by path only and does not set or restrict Host; production overlay is not in the repository",
          "evidence_file": null
        }
      ],
      "fingerprints": [
        "identity:reset:host-derived-link"
      ],
      "unresolved": [
        "whether the production proxy overrides or validates Host"
      ],
      "attempts": []
    },
    {
      "coverage_id": "cu-b21b2524c4ae7d41843c",
      "refs": {
        "surface": "src/api/webhooks.py#POST /webhooks/test",
        "boundary": "src/net/fetch.py#fetch_url",
        "subsystem": "services/integrations",
        "attack_class": "attack-classes.md#Files, fetches, and resource handling"
      },
      "label": "Webhook test: server-side fetch of a caller-supplied URL",
      "starting_paths": [
        "src/api/webhooks.py",
        "src/net/fetch.py"
      ],
      "class_blocks": [
        "attack-classes.md#Files, fetches, and resource handling"
      ],
      "excluded_blocks": [],
      "origin": "none",
      "wave": 1,
      "status": "blocked",
      "owner": "hunter-07",
      "reviewed_paths": [
        "src/api/webhooks.py",
        "src/net/fetch.py"
      ],
      "checks": [
        {
          "agent_id": "hunter-07",
          "reviewed_paths": [
            "src/api/webhooks.py",
            "src/net/fetch.py"
          ],
          "invariant": "The server refuses to fetch internal addresses, including after a redirect",
          "method": "source",
          "result": "an allowlist check runs on the first URL; redirect handling is delegated to a client wrapper whose configuration is loaded at run time",
          "evidence_file": null
        }
      ],
      "fingerprints": [],
      "unresolved": [
        "redirect policy is set by a runtime config file not present in the repository and no sandbox was available to observe it"
      ],
      "attempts": []
    },
    {
      "coverage_id": "cu-b85a24022c50ffa86b8e",
      "refs": {
        "surface": "web/src/App.tsx#router",
        "boundary": "web/src/render/markdown.tsx#renderNote",
        "subsystem": "web",
        "attack_class": "attack-classes.md#Injection"
      },
      "label": "Web client: rendering of user-written order notes",
      "starting_paths": [
        "web/src/render/markdown.tsx"
      ],
      "class_blocks": [
        "attack-classes.md#Injection"
      ],
      "excluded_blocks": [],
      "origin": "none",
      "wave": 1,
      "status": "out_of_scope",
      "owner": null,
      "reviewed_paths": [],
      "checks": [],
      "fingerprints": [],
      "unresolved": [
        "companion_outside_tier:client-side.md"
      ],
      "attempts": []
    },
    {
      "coverage_id": "cu-c565fdde20a55cc15828",
      "refs": {
        "surface": "src/api/orders.py#POST /orders/{id}/refund",
        "boundary": "src/authz/policy.py#require_owner",
        "subsystem": "services/orders",
        "attack_class": "attack-classes.md#Access control"
      },
      "label": "Refund endpoint: owner check on order refunds",
      "starting_paths": [
        "src/api/orders.py",
        "src/authz/policy.py"
      ],
      "class_blocks": [
        "attack-classes.md#Access control"
      ],
      "excluded_blocks": [
        {
          "block": "attack-classes.md#Injection",
          "reason": "handler takes a numeric path ID and a fixed enum reason; no interpreted sink"
        }
      ],
      "origin": "none",
      "wave": 1,
      "status": "candidate",
      "owner": "hunter-01",
      "reviewed_paths": [
        "src/api/orders.py",
        "src/authz/policy.py",
        "src/db/orders.py",
        "src/services/refunds.py",
        "tests/api/test_refunds.py"
      ],
      "checks": [
        {
          "agent_id": "hunter-01",
          "reviewed_paths": [
            "src/api/orders.py",
            "src/services/refunds.py",
            "src/db/orders.py"
          ],
          "invariant": "A refund may only be issued by the tenant that owns the order",
          "method": "source",
          "result": "refund handler loads the order by ID alone; require_owner is applied to GET and PATCH but not to the refund route",
          "evidence_file": null
        },
        {
          "agent_id": "hunter-01",
          "reviewed_paths": [
            "tests/api/test_refunds.py"
          ],
          "invariant": "Tenant B cannot refund tenant A's order",
          "method": "local",
          "result": "existing test client with two fixture tenants: tenant B's request returned 200 and the fixture order moved to refunded",
          "evidence_file": "agents/hunter-01/evidence/refund_cross_tenant.txt"
        },
        {
          "agent_id": "verifier-01",
          "reviewed_paths": [
            "src/api/orders.py",
            "src/authz/policy.py",
            "src/services/refunds.py"
          ],
          "invariant": "No middleware or service-layer check enforces ownership on the refund path",
          "method": "source",
          "result": "re-read the route table and service; no ownership check exists between the handler and the write",
          "evidence_file": null
        }
      ],
      "fingerprints": [
        "orders:refund:missing-owner-check"
      ],
      "unresolved": [],
      "attempts": []
    }
  ]
}
rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first.Stars: 194kUpdated Claude CodeCodexVerifiedInstall