Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 5 additions & 1 deletion .claude/skills/create-fetcher/SKILL.md
Original file line number Diff line number Diff line change
Expand Up @@ -140,7 +140,11 @@ Read these before building (skim, don't summarize back):
`docs/authoring_fetcher_guide.md` ("The `description` and `instructions` text"): a
one-sentence description, then plain-text instructions in a `|` block with
Script (`<category>/<short_name>/<entry>`), Access, Commands Executed, Output
and Summary fields. Paramify shows them as typed, so no markdown.
and Summary fields. Paramify shows them as typed, so no markdown. Hard limits:
the description is one sentence of at most 30 words ending in a period, with
no fetcher name; a summary field line is at most 800 characters; the
instructions at most 20,000. `tests/test_contracts.py` checks every new
fetcher's text.
- `category` — from the tool name.

**Ask the user (batch into ONE `AskUserQuestion`, ≤3 questions):**
Expand Down
5 changes: 4 additions & 1 deletion .claude/skills/onboard-platform/SKILL.md
Original file line number Diff line number Diff line change
Expand Up @@ -101,7 +101,10 @@ person. The conventions it reads are in `references/state.md`.
`teardown.sh`. The derivation of each field goes in the fetcher's
`evidence_set.instructions` (laid out as in `docs/authoring_fetcher_guide.md`:
Script, Access, Commands Executed, Output, Summary fields), where an assessor
reads it; the build story goes
reads it. Hard limits: the description is one sentence of at most 30 words
ending in a period, with no fetcher name; no markdown; a summary field line
at most 800 characters; the instructions at most 20,000. The contract test
checks every new fetcher's text. The build story goes
in `notes/`. The third run wrote 45–74-line module docstrings restating both.
Committed code never cites `.onboarding/`, which is gitignored.
- **In the public repo, the branch carries fetchers, never new validators.**
Expand Down
8 changes: 5 additions & 3 deletions docs/authoring_fetcher_guide.md
Original file line number Diff line number Diff line change
Expand Up @@ -163,16 +163,18 @@ It fails on:

[`azure/alert_rules`](../fetchers/azure/alert_rules/fetcher.yaml) is a full
example. `tests/test_contracts.py` checks:
- the layout, for the fetchers listed in `TEXT_FORMAT_FETCHERS`;
- the layout, for every fetcher except those on `TEXT_FORMAT_PENDING` (older
fetchers that still fail it; the list only shrinks, and a new fetcher is
never added to it);
- that a test calls `assert_summary_documented` for every fetcher whose
instructions list summary fields. Fetchers that listed them before the check
existed wait in `SUMMARY_DOCS_PENDING`, which is meant only to shrink. Its
count, `SUMMARY_DOCS_PENDING_MAX`, must match it, so adding a fetcher means
raising the count, which shows in review. A new fetcher calls the fixture
instead.

Until a platform is converted, its instructions only need to open with the
`Script:` path.
A fetcher on `TEXT_FORMAT_PENDING` only needs instructions that open with the
`Script:` path, until its text is rewritten and it leaves the list.

## 3. Write `fetcher.py` and smoke-test it

Expand Down
159 changes: 94 additions & 65 deletions fetchers/azure/ai_services_configuration/fetcher.yaml

Large diffs are not rendered by default.

16 changes: 4 additions & 12 deletions fetchers/azure/entra_app_registrations/fetcher.yaml
Original file line number Diff line number Diff line change
@@ -1,14 +1,6 @@
name: azure_entra_app_registrations
version: 0.2.1
description: >
Every Microsoft Entra ID app registration with each of its client secrets and
certificate credentials and that credential's expiry date, so credential rotation
is evidenceable per credential. Three expiry states are kept distinct — already
expired, inside the 30-day rotation window, and no expiry date at all — because a
never-expiring credential never appears in an "expiring soon" report by
construction. Rotation compliance is reported over credentials, excluding apps that
hold none. Each credential also carries the lifetime it was issued for, its end
date minus its start date.
description: Lists every Microsoft Entra ID app registration with each client secret and certificate, its expiry state and issued lifetime, so credential rotation can be shown per credential.
category: azure

supports_targets: true
Expand Down Expand Up @@ -96,14 +88,14 @@ evidence_set:

Do credentials expire and get rotated?
- total_credentials — Number of secrets and certificates across all app registrations; null if the applications read failed or stopped part-way
- password_credentials — Number of those that are client secrets; null if the applications read failed or stopped part-way
- certificate_credentials — Number of those that are certificates; null if the applications read failed or stopped part-way
- password_credentials — Number of credentials that are client secrets; null if the applications read failed or stopped part-way
- certificate_credentials — Number of credentials that are certificates; null if the applications read failed or stopped part-way
- expired_credentials — Number of credentials whose end date has passed; null if the applications read failed or stopped part-way
- credentials_expiring_soon — Number of credentials that expire within rotation_threshold_days days and haven't expired; null if the applications read failed or stopped part-way
- credentials_without_expiry — Number of credentials with no end date; null if the applications read failed or stopped part-way
- healthy_credentials — Number of credentials that have an end date, haven't expired and don't expire within rotation_threshold_days days; null if the applications read failed or stopped part-way
- rotation_threshold_days — Days before expiry at which a credential counts as expiring soon (30)
- credential_rotation_compliance_percentage — Percentage of credentials that are healthy; 0 when there are none; null if the applications read failed or stopped part-way
- credential_rotation_compliance_percentage — Percentage of credentials that are healthy = healthy_credentials / total_credentials × 100 (a whole number, 0–100); 0 when total_credentials is 0; null if the applications read failed or stopped part-way
- apps_with_expired_credentials — Number of app registrations holding an expired credential; null if the applications read failed or stopped part-way
- apps_with_credentials_expiring_soon — Number of app registrations holding a credential expiring soon; null if the applications read failed or stopped part-way
- apps_with_credentials_without_expiry — Number of app registrations holding a credential with no end date; null if the applications read failed or stopped part-way
Expand Down
3 changes: 1 addition & 2 deletions fetchers/azure/entra_sign_in_logs/fetcher.yaml
Original file line number Diff line number Diff line change
@@ -1,7 +1,6 @@
name: azure_entra_sign_in_logs
version: 0.3.0
description: >-
Shows from the workspace's Entra sign-in logs whether successful sign-ins required MFA and phishing-resistant methods, the changes made to the authentication and Conditional Access policies, and, as counts only, the non-interactive user sign-ins and the service-principal (application) sign-ins that detections and least-privilege reviews depend on.
description: Shows from Entra sign-in and audit logs whether successful sign-ins required MFA and phishing-resistant methods, what changed in authentication and Conditional Access policies, and counts of non-interactive and service-principal sign-ins.
category: azure

supports_targets: true
Expand Down
5 changes: 5 additions & 0 deletions tests/test_azure_resource_exclusions_ai.py
Original file line number Diff line number Diff line change
Expand Up @@ -256,6 +256,11 @@ def test_services_without_patterns_adds_only_empty_exclusion_fields(monkeypatch,
}


def test_services_summary_matches_the_instructions(monkeypatch, tmp_path, assert_summary_documented):
_, evidence, *_ = run_services(monkeypatch, tmp_path, "rg:rg-out")
assert_summary_documented("azure/ai_services_configuration", evidence["summary"])


def test_services_exclude_by_group_and_region_and_skip_account_reads(monkeypatch, tmp_path):
code, evidence, calls, paths, diag = run_services(monkeypatch, tmp_path, "rg:rg-out, location:regioneast, rg:typo")
summary, results = evidence["summary"], evidence["results"]
Expand Down
Loading
Loading