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
2 changes: 1 addition & 1 deletion CLAUDE.md
Original file line number Diff line number Diff line change
Expand Up @@ -725,7 +725,7 @@ project's recurring failure family arriving through process rather than code.
| **`risk` scoring** | Samika | Done (#60) — `POST_PROCESSORS`, ruleset in `docs/severity-rules.md`. Re-rates severity and sorts worst-first; the two limits are enforced in `pipeline.run_post_processors()`, not trusted |
| Dashboard + secure upload | Samika | Done — real findings on screen since #39 |
| PF Sense conversion | Ankeet | Done — `analysis/pfsense_convert.py`, and **hardened**: refuses config injection and path traversal via free-text fields (#53), an unbound ACL / empty rule set / unvalidated addressing (#54), and ambiguous rule order (#58, closing #47). Also refuses an interface tag that is not a plain name (a namespaced tag injected config lines), and, per rule, `<not/>`, IPv6-only and ICMP-type rules; disabled rules are left out, as pfSense does. See §7 |
| **`change_impact`** | Shubham | Done (#140, merged 17 August) — `analysis/change_impact.py`, `analyse_change(before, after)`. Shape C: **not** in `CHECKS`, and both the module docstring and the registry comment say so. Pairs `compareFilters` (which ACL *lines* moved) with `differentialReachability` (which *traffic* changed fate), because a changed line that moves no traffic is noise and moved traffic with no changed line is the case a filter diff alone misses. Direction-aware per `docs/policy-rules.md`: opening `high`, tightening `medium`, both reported |
| **`change_impact`** | Shubham | Done (#140, merged 17 August) — `analysis/change_impact.py`, `analyse_change(before, after)`. Shape C: **not** in `CHECKS`, and both the module docstring and the registry comment say so. Pairs `compareFilters` (which ACL *lines* moved) with `differentialReachability` (which *traffic* changed fate), because a changed line that moves no traffic is noise and moved traffic with no changed line is the case a filter diff alone misses. Direction-aware per `docs/policy-rules.md`: opening `high`, tightening `medium`, both reported. Since #389 (closing #323, merged 14 October) it also runs every `policy_compliance` rule before and after and reports which the change breaks, fixes or alters; since #403/#402 it reports devices added or removed, treats rules about a device in neither config as not applicable, and says which policy sections it did **not** compare |
| **A user policy reaching a check** | Arsh | Done (#181, merged 28 August) — `analysis/policy.py` loads and validates it; `policy_compliance` reads it via `rules_in_use()`, keyed on `active_policy() is None` rather than on emptiness. Measured below: policy-driven detections on a stranger's network go 3 → 8 |
| **Draft a policy from your own config** | Arsh | Done (#326) — `analysis/draft_policy.py`, `/api/policy/draft`, a button under the policy picker. One `policy_compliance` rule per line of each applied filter, citing the file and line Batfish records for it. Each rule is checked with the same `searchFilters` question the check will ask, so a line an earlier line partly decides is listed as not drafted, with Batfish's example packet, rather than copied. `kind` and `violation_severity` are left as `<FILL IN ...>`, and `analysis/policy.py` refuses the file until they are decided and its `draft` block is deleted: accepted as it stands, a draft passes on `rtr-us5-insecure` |
| **What Netwise could not read** | Arsh | Done (#317) — `analysis/parse_coverage.py`. A partly parsed file's "Results may be incomplete" finding now says how many configuration lines Batfish did not recognise, by kind and line number, from Batfish's own `parseWarning`. On `vendor-asa`: 5 of 16, including the `access-group` line behind #373. Comment and blank lines are left out of the count only where the format's marker is known; "31 of 36" would have overstated it. Lives in the finding's detail, so it reaches every surface without an F-1 change |
Expand Down
Loading
Loading