Address adversarial-review findings on /unlink:
- Self-service-in-DM / cross-guild destructive (CRITICAL): the MANAGE_ROLES gate is not
trusted for auth -- poise 0.6 treats a DM invoker as Permissions::all(), so it passed
in a DM and let /unlink @self delete the row before any guild check. The command is now
guild_only and, as the FIRST thing before any DB op, asserts ctx.guild_id() is the DSEC
guild specifically (which also stops a moderator of another guild the bot is in from
deleting DSEC rows). The gate is kept for command visibility only.
- Expired-interaction mutate-then-skip-audit (MED): defer_ephemeral is sent before any
DB/HTTP work; the moderator reply is best-effort so a failed reply cannot skip the audit
log, which now always runs.
- Delete failure skipped audit + could leak (MED): the row delete no longer uses `?`. A
delete error is handled in-command with a sanitized ephemeral (no raw error) and a
distinct failure audit line; the raw error is redacted (redact_digits) before it reaches
stderr.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017XrE7F9ZuBWdQnS8CZvYDE
Address adversarial-review findings on the verify path:
- Already-linked caller bypass (HIGH): add_dsec_discord_table no longer early-returns
Linked on any existing row. It now checks the SUBMITTED id's owner first (idempotent
only when that owner is the caller) and refuses a caller already linked to a DIFFERENT
student id -- closing the path where a stale COR-03 partial-insert row let an account
claim someone else's id.
- Check-then-insert race (HIGH): the insert now catches a UNIQUE(student_id) violation
(SQLSTATE 23505), re-queries the owner and converts it to the generic refusal + audit.
SECURITY.md now marks the live UNIQUE constraint (after the dup sweep) as a merge/deploy
gate, not optional.
- PII leak in error prints (HIGH): raw supabase/reqwest errors (which embed
student_id=eq.<id> and a 23505 Key detail) are never printed on the verify path. New
redact_digits() masks 7+ digit runs; the interaction id is used as an opaque
correlation ref. student_id_owner now selects only discord_id.
- Concurrent-attempt rate-limit bypass (HIGH): a per-UserId tokio Mutex (AppState.
verify_locks) serializes a whole verification attempt so concurrent modal submits
cannot each slip under the 5-in-15min limit. Requires tokio "sync" feature.
- Ownership conflict never counted (MED): a refused link now records exactly one failed
attempt (success/infra-error record none), so a stolen-but-claimed credential can no
longer loop the query set forever.
- Expired-interaction mutate-then-skip (MED): defer_ephemeral is now sent the instant the
modal arrives, before any DB/logs work; every later reply edits the deferred response.
- Log oracle (MED): the failed-attempt audit line is one fixed generic string, so
"not found" vs "name mismatch" are indistinguishable in the logs channel too.
Also: on_error no longer forwards a handled FrameworkError::Command to poise's builtin
(whose Command arm does a non-ephemeral ctx.say(raw_error), leaking DB error text); mutex
locks recover from poisoning instead of panicking; and the attempt/lock maps evict
stale/idle entries so they cannot grow unbounded.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017XrE7F9ZuBWdQnS8CZvYDE
Add a moderator-gated /unlink <user> slash command (required_permissions =
"MANAGE_ROLES", same gate as embed()). It deletes the dsec_discord_members row FIRST,
then removes the verified role -- that order avoids leaving a member un-roled but still
linked, the state that permanently breaks /member_info. The delete uses .returning(...)
so PostgREST returns the row body (the COR-03 empty-204 gotcha) and so we can tell
whether a link actually existed. It replies ephemerally with what it did and reports
clearly if the role-removal half fails so a human can finish it, then writes a log_embed
entry to the logs channel naming the acting moderator. Registered in main.rs.
Add SECURITY.md: the moderator runbook for undoing a link (via /unlink and by hand),
who holds the Supabase credentials, the fact that deleting the row does NOT revoke the
Discord role (why /unlink does both), and the owner-only SEC-19 UNIQUE-constraint step.
A member-facing /unverify is deliberately not added. The /member_info leave/trim/remove
policy call is an owner decision and is intentionally not implemented here.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017XrE7F9ZuBWdQnS8CZvYDE
Collapse both verification failure paths into one shared verification_failed_embed()
constructor so "not found" and "name mismatch" are byte-identical and can no longer be
used as an oracle for whether a student id holds an Active membership.
Remove the stale in-memory student_cache entirely: it was consulted before the only
query carrying membership_status = "Active", had no TTL and was never evicted, making it
a stale-membership bypass in a long-lived container. fetch_student now runs on every
verification attempt before any role grant.
Add a per-Discord-user failed-attempt counter in AppState
(Mutex<HashMap<UserId, (u32, Instant)>>). After 5 failures in 15 minutes an attempt is
refused with the generic embed and no database round trip. Each failure is logged to the
logs channel with the Discord user id and a timestamp only -- never the submitted name
or student id.
Add an application-level student-id uniqueness check in add_dsec_discord_table: if a
different discord_id already holds the student id, refuse with the generic embed and log
the conflict to the logs channel with both discord ids (not the student id).
The UNIQUE constraint DDL and duplicate sweep on live Supabase are an owner step (see
SECURITY.md); the emailed-OTP possession proof is noted as a follow-up TODO.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017XrE7F9ZuBWdQnS8CZvYDE
* COL-BOT-03: make /member_info expiry reply ephemeral and drop dead query
Restructure the membership lookup to `let Some(user_data) = ... else { ... }`
so the "Couldn't find info" branch now sends `.ephemeral(true)` like the other
two replies; whether a member's membership has lapsed is no longer announced to
the whole channel. Delete the commented-out `student_data` query that duplicated
member_data(). No embed fields or wording changed.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017XrE7F9ZuBWdQnS8CZvYDE
* COL-BOT-05: parse all config once in AppState, remove per-event env reads
Finishes the AppState config hoisting (COR-12's groundwork was not present in
the repo, so this ticket establishes it): AppState now carries guild_id,
honeypot_channel_id, leetcode_channel_id, verified_role_id, logs_channel_id and
weather_token, all parsed once in AppState::new(). A missing or unparseable id
now produces one boot error naming every offending variable plus a pointer to
.env.example, instead of a per-event .expect() panic (a data race under the
multi-threaded runtime, since dotenv() calls the now-unsafe set_var).
- main.rs: new fields, required_u64() collector, fail-fast with combined message.
- events/message.rs: honeypot/create_leetcode_thread/on_message take &Data and
read channel/guild ids from AppState; dropped dotenv() and env::var.
- events/interaction_create.rs: deleted verified_role_id(); use
data.state.verified_role_id.
- commands/mods_only.rs: log_embed takes logs_channel_id: ChannelId; caller in
message.rs passes data.state.logs_channel_id.
- commands/weather.rs: WEATHER_TOKEN read once into AppState (non-fatal), passed
to get_weather — required so the new CI grep guard can be clean.
- .env.example: add LEETCODE_CHANNEL_ID.
- .github/workflows/ci.yml: add "No environment reads outside main.rs" guard.
grep -rn 'dotenv()\|env::var' src/ | grep -v '^src/main.rs:' now returns nothing.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017XrE7F9ZuBWdQnS8CZvYDE
* COL-BOT-02: guard the honeypot, log failed bans, stop log_embed panicking
The honeypot banned whoever posted, unconditionally. Add guards so it never
bans a bot, a webhook, the bot itself, or anyone who can moderate
(ban_members / manage_messages / administrator, computed from the cached guild).
A failed ban now logs the user and error instead of being silent.
Remove the .expect("LOG FAIL") in log_embed: it runs inside a message handler,
so a transient Discord failure was a panic; it now eprintln!s and returns.
Decouple on_message so honeypot and the LeetCode thread each run and log
independently — a failure in one no longer skips the other, and neither
propagates out of the event handler.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017XrE7F9ZuBWdQnS8CZvYDE
* COL-BOT-04: clamp LeetCode thread names and add the repo's first tests
create_title_from_message dropped the two-character numbering prefix but a
first line of two chars or fewer left an empty name, and a line over 100 chars
exceeded Discord's limit — both produced a silent 400. Move the function to
module level (&str), clamp to 100 characters (on a char boundary, not bytes),
and fall back to "LeetCode discussion" when nothing is left. A failed
create_thread now logs instead of propagating. Add six unit tests covering
normal input, empty, a two-char line, a >100-char line, a multibyte first
character, and a multi-line message — the first tests in this repo.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017XrE7F9ZuBWdQnS8CZvYDE
* COR-03: fix verification failing on a member's first attempt
The insert into dsec_discord_members deserialised the response into
Vec<DiscordMemberRow>, but without `.returning(...)` supabase-lib-rs never sends
`Prefer: return=representation`, so PostgREST returns a 201 with an empty body.
Parsing that empty body failed and the `?` aborted before add_role and before
any reply — so every member's first click died with "This interaction failed",
while the row was still written (which is why the second attempt worked). Add
`.returning("student_id,discord_id")` so the row comes back and parses.
Also wrap the post-modal verification work so any failure sends the user an
ephemeral "something went wrong" instead of a dead interaction (poise's on_error
has no handle to the modal submission), and set an explicit on_error on
FrameworkOptions that replies ephemerally on command errors and keeps the
default logging. The database-write-first ordering is left unchanged.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017XrE7F9ZuBWdQnS8CZvYDE
* NEW-UXA11Y-11: normalise names and student IDs before comparing them
Verification compared names with exact lowercased equality and never trimmed, so
a trailing space, a double space, or a missing middle name turned a real paid
member away — and a pasted "s123456789 " was looked up with the trailing space
and missed. Add normalise_name (trim + collapse whitespace), normalise_student_id
(strip whitespace and a leading "s"), and name_matches (equal, or every typed
word present in the roster name in order), and route the database path, the cache
path (cached_name_matches / cache_student) and the student-id lookup through them.
"Doe John" still does not match "John Doe" and an empty name matches nothing.
Update the modal placeholder to "As it appears on your DUSA membership". Four unit
tests added. The two failure embeds are left byte-identical (SEC-19 owns those).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017XrE7F9ZuBWdQnS8CZvYDE
* OPS-04: restart on crash, real logging, and a deploy that fails when the bot does
- docker-compose.yml: `restart: unless-stopped` so a panic, bad frame or VPS
reboot brings the bot back instead of leaving it dead. Comment explaining why
no healthcheck (slim runtime, no HTTP port).
- Cargo.toml: enable tracing-subscriber's env-filter feature (declared but never
initialised until now).
- main.rs: initialise tracing as the first statement in main(), defaulting to
`info` — serenity/poise/supabase logs now surface, without leaking student IDs
or the service email that the Supabase client emits at `debug`.
- deploy.yml: after `up -d`, assert the container is actually Running 30s later
and dump its logs and fail if not — `up -d` returns 0 on container creation,
not on a working process.
- Dockerfile: pin the builder to rust:1-bookworm to match the bookworm-slim
runtime, removing the trixie/bookworm glibc mismatch.
- README: document the info-level default and the RUST_LOG=debug PII footgun.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017XrE7F9ZuBWdQnS8CZvYDE
* NEW-UXA11Y-11: require first AND last name to match (close a verification bypass)
Codex review: name_matches accepted any non-empty ordered subset, so a single
common token ("John" or "Doe") matched "John Michael Doe" — a student id plus one
name token could claim the verified role for someone else. Now both the first and
last tokens must match, middle tokens the student typed must appear in the roster
in order (omitted middles still fine), and a single token, an arbitrary subset,
or a wrong first/last name is rejected. Add tests: single-token input, wrong
surname and wrong first name do NOT verify; full name and first+last with the
middle omitted do.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017XrE7F9ZuBWdQnS8CZvYDE
* OPS-04: make the deploy assertion catch a crash-looping bot
Codex review: with `restart: unless-stopped` a bot that panics at boot is
restarted, so a single `docker inspect` still reads Running while it crash-loops
— the deploy went green while broken. The assertion now waits for the bot to
reach READY (it logs "Logged in as ..." from the Ready handler) within a
stability window AND requires RestartCount == 0, breaking early and failing the
moment a restart is observed, dumping container logs on any failure.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017XrE7F9ZuBWdQnS8CZvYDE
* COL-BOT-05: stop /weather panicking on missing token or error payloads
Codex review: a missing/blank WEATHER_TOKEN and invalid locations produced an API
error body that then hit as_str().unwrap() calls. Store the token as
Option<String> (None when missing or blank) and disable /weather with a friendly
ephemeral reply instead of running with an empty key. get_weather now returns the
HTTP status with the parsed body; a non-2xx surfaces the API's error message, and
every success field is read via JSON pointers with no unwrap() on external JSON.
The thumbnail is only set when an icon URL is actually present.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017XrE7F9ZuBWdQnS8CZvYDE
* COL-BOT-02: make the honeypot moderator guard fail safe
Codex review: the guard fell THROUGH to banning when the member lookup or the
permission calculation failed, and it ignored per-channel permission overwrites.
Now it computes effective permissions in the honeypot channel (honouring
overwrites via Guild::user_permissions_in) and ABSTAINS from banning whenever the
guild, the member, or the channel cannot be resolved — fail-safe: never ban when
identity or permissions cannot be established. Replaces the deprecated
Member::permissions call, so the #[allow(deprecated)] is gone.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017XrE7F9ZuBWdQnS8CZvYDE
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
* SEC-01: add pull-request CI and stop the deploy leaving DOT_ENV on the VPS
A merge to main runs the merged commit as root on the club VPS through the
self-hosted runner, and this repo has no PR CI, no required review and no
status check.
- .github/workflows/ci.yml: fmt / clippy / build --locked / test on
ubuntu-latest, contents: read, both actions pinned by commit SHA. Never
the VPS.
- deploy.yml: keep push to main, add workflow_dispatch, permissions
contents: read, environment: production, and an if: always() step that
removes the .env the deploy writes into the workspace.
- .dockerignore: keep that .env, target/ and .git out of the build context.
- Dockerfile: --locked on both cargo build --release lines.
environment: production gates nothing until required reviewers are configured
on the environment itself, and the ruleset still requires zero approving
reviews. Both are owner-only and listed in the pull request.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* SEC-01: make cargo fmt and clippy clean so the CI job can go green
The CI workflow added in #5 fails on code that predates it: three files were
unformatted and `cargo clippy --all-targets -- -D warnings` reported 26 errors.
A red check cannot be made a required status check on the protect-main ruleset,
which is what this unblocks.
`cargo fmt --all` over three files, and 26 clippy errors resolved: 18 via
`cargo clippy --all-targets --fix`, the rest by hand. Two fixes uncovered lints
that had been masked (an `unnecessary_unwrap` in info.rs behind the needless
borrow on the line above, and two `unnecessary_to_owned` at the call sites of a
signature that moved from `&String` to `&str`), so 28 fixes for 26 warnings.
Three `#[allow]`s where the only real fix would change a signature or a public
API: `result_large_err` on `AppState::new` (the large variant is
`supabase::Error`, owned by supabase-lib-rs) and `too_many_arguments` on
`log_embed` and on the `embed` slash command.
No behaviour change. src/ only; .github/, Dockerfile, .dockerignore, Cargo.toml
and Cargo.lock are untouched.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* SEC-01: pin actions/checkout in the deploy workflow to a commit SHA
deploy.yml runs on the self-hosted VPS runner with passwordless sudo, so a
mutable tag on this action is a code-execution path onto that host if the tag
is ever moved. Pinned to the commit v4 currently resolves to
(11d5960a326750d5838078e36cf38b85af677262), verified against upstream — this is
the same code the deploy already runs today, not a version bump. ci.yml is on
v7.0.1; the deploy path is deliberately left on v4 so that pinning does not
smuggle a major-version change into a workflow whose only test is a live deploy.
Raised by Codex review of PR #5 as the one unpinned `uses:` in either workflow.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
