* fix(agent): filter skill slash commands by the agent's grants
The inline <available_skills> block was already filtered by visibility and
agent grants, but slash activation read the loader's full skill list. A
/<slug> command could therefore activate a skill the agent was never granted,
and the not-found suggestions disclosed that such skills existed.
Resolve slash commands against the agent's allow list via FilterSkills. The
list is filtered before matching, so /list-skills, /help and the near-match
suggestions are all covered by the same change.
The allow-list convention is unchanged: nil means every skill (the fallback
when the access store errors), an empty slice means none, and a populated
slice is an explicit set of slugs.
* fix(agent): split slash commands on any whitespace
The tokenizer separated the skill name from the rest of the message with
strings.Cut(after, " ") — a literal space. A user who typed the command and
pressed Enter before the rest of the message sent "/ck:git\nreview the diff",
which parsed to the target "ck:git\nreview" and matched no skill. The reply
was "skill not found" followed by near-matches that included the skill they
had just named, because the similarity fallback searched the mangled string
and still landed beside it. Multi-line messages are ordinary in Slack and
Telegram, so this was reachable in normal use.
Introduce cutFirstField, which splits on the first run of whitespace, and use
it everywhere the literal-space split appeared. The match loop had the same
assumption in strings.HasPrefix(raw, value+" "); hasFieldPrefix replaces it
and decodes the following rune properly, so a multi-byte skill name is not
truncated. That also stops a plain string prefix from matching: a command for
"frontend-design-extra" no longer resolves to "frontend-design".
* fix(tools): inherit the parent agent's tool policy in spawned subagents
buildSubagentToolsRegistry clones the parent registry, then overwrites exec,
read_file, write_file and list_files with freshly constructed tools. A fresh
tool carries none of the hardening the gateway applies to the parent's
instances at startup: exec path denials and their exemptions, the shell
deny-group toggles, the command keyword allowlist, and the read/write/list
deny prefixes covering config.json, the internal databases and delegate/.
Spawning a subagent therefore widened what an agent could reach — the parent
was blocked from the data dir, the subagent was not. Verified by disabling
the new call: the subagent's exec read config.json out of the denied data
dir, read_file did the same, and write_file overwrote it.
Copy the policy from the live parent instances rather than re-deriving it, so
there is one source of truth and a deny path reloaded later through config
pub/sub reaches subagents without a second wiring site that can drift.
Allow-prefixes are inherited alongside the denials: they are what make the
denied roots usable at all, since the skills store sits under the denied data
dir. Copying denials without them would leave a subagent unable to read the
skills it is told to use.
* fix(agent): use one inline-vs-search decision for prompt and preview
The system-prompt preview exists to show the prompt an agent actually gets,
but it decided between inline skills and search mode on its own terms: it
counted tokens with the fallback counter over the fully rendered XML — tags
and <location> paths included, at roughly runes/2 — while the prompt builder
estimated name+description at chars/4. On the same eighteen skills that read
3087 against 944, so the preview reported search mode for an agent that was
running inline.
Extract shouldInlineSkills and call it from both paths. The estimate keeps
the prompt builder's rule, mirroring BuildSummary's 200-rune description
truncation, so the decision still costs no rendering.
The preview's loader interface gains FilterSkills to feed it. Its behaviour
on an access-store error is unchanged: an empty allow list yields no skills
rather than falling back to showing every skill.
* fix(http): keep reference files when importing skills
Export archives the whole skill directory, but import recognised only
metadata.json, SKILL.md and grants.jsonl. The switch had no default, so every
other file was discarded without a log line. A skill whose SKILL.md cites
references/*.md arrived without them, the import reported success, and the
skill failed at the first read.
Collect the remaining entries and write them under the skill directory with
their structure intact. Archive entry names are attacker-controlled, so each
path goes through sanitizeRelPath: a traversal attempt collapses to a
relative path and the write stays inside the skill directory.
* test(agent): make the inline-decision table exercise both gates
The "100 skills, 200-char descriptions" case claimed to cover the token
ceiling, but 100 exceeds the count ceiling so the count check short-circuited
and the token branch never ran — deleting that branch would not have failed
the test.
Split the table so each case isolates one gate: tiny descriptions for the
count rows, a count safely under the ceiling for the token rows. Removing the
token comparison now fails the over-the-ceiling case and nothing else.
* fix(agent): gate slash commands on the managed tier only
The allow list comes from a query over the `skills` table, so filesystem-tier
skills — the workspace, .agents and ~/.agents directories of the five-tier
loader — have no row in it. Filtering every skill against the list made those
four tiers unreachable by slash while skill_search still found them
unfiltered, which turned a security fix into a functional regression.
Apply the list to managed skills only. Builtins are seeded with is_system and
returned unconditionally, so they stay reachable either way.
This closes slash activation of ungranted managed skills. It does not close
skill_search and use_skill, which still read the loader's full list; that is a
wider change and wants its own review.
* fix(http): stop imported skill files from replacing guarded ones
Two ways the auxiliary write could go wrong.
GuardSkillContent scans SKILL.md before anything reaches disk, but the switch
that routes archive entries compares the raw path. An entry named
"./SKILL.md" does not match it, so it fell through to the auxiliary set —
where sanitizeRelPath collapses it back to "SKILL.md" and the write replaced
the file that had just been scanned. Reject any auxiliary path that resolves
to a name the import handles by itself, and log the attempt.
Tar directory entries carry a trailing separator and no content. Written as
files they took the name a real directory needed, so everything beneath was
dropped — and whether that happened depended on map iteration order, making
it intermittent. Skip them; MkdirAll creates what is needed.
Also cap the number of auxiliary files per skill and log when the cap trims
an archive, so a truncated import is visible rather than silent.
---------
Co-authored-by: mor-phongdt <phong.dangtuan@mor.com.vn>
- Give Name/Status/Tokens/Spans/Time explicit widths so Name cannot shove others off-screen
- Show status label next to the status icon
Fixes#762
Signed-off-by: Vedant Madane <6527493+VedantMadane@users.noreply.github.com>
Editing the description in onboarding step 3 deselected the preset, and the
display name, agent key and emoji were derived from that selection. Losing it
dropped them to a hardcoded "Fox Spirit" / "fox-spirit" / fox emoji, so picking
the Writer preset and rewriting its prompt created an agent named Fox Spirit.
The step also offered no way to name the agent at all.
Identity is now editable state that a preset only prefills:
- Add display name, emoji and agent key inputs to both onboarding steps, with
the key auto-derived from the name until it is edited by hand.
- Editing the prompt only clears the preset highlight; the identity stays.
- Block submit on an empty name or an invalid slug instead of falling back.
- Give the web presets stable English agent keys, so the slug no longer depends
on the UI locale, and strip the emoji prefix off the preset label before it
becomes a display name.
- While a preset is active and its field is untouched, the name and prompt still
follow a locale switch.
Reuses the existing agents:create.* keys, present in en/vi/zh on both surfaces.
DeepSeek rejects a request whose assistant history omits reasoning_content
once the conversation is in thinking mode:
HTTP 400: The `reasoning_content` in the thinking mode must be passed
back to the API.
A thinking-mode turn that only emits tool calls produces no reasoning of
its own, so Message.Thinking is empty and the wire field was dropped. The
next request then mixes assistant messages with and without the field and
upstream fails. Only kimi_coding had an empty-string fallback.
Detect thinking mode from the conversation itself: when any assistant
message carries reasoning, every replayed assistant message must carry the
field, empty string included. Narrower than matching on model name, which
would also inject the key for non-thinking models such as deepseek-chat.
Reproduces from a run where iteration 1 returned reasoning plus an edit
call, the edit failed, iteration 2 returned a bare read_file call, and
iteration 3 was rejected.
ExtractWikilinks builds a ±25-byte context window around each wikilink
match. When the content contains multi-byte UTF-8 characters (Vietnamese,
CJK, emoji), the byte-offset start/end can fall on a continuation byte,
producing invalid UTF-8 that PostgreSQL rejects with SQLSTATE 22021
('invalid byte sequence for encoding UTF8').
Fix: wrap the byte-offset slice with strings.ToValidUTF8, which strips
any leading/trailing partial runes. This is a minimal one-line change
that preserves the existing context semantics while ensuring the output
is always valid UTF-8.
Available since Go 1.21 (this project uses Go 1.26).
Closes#1472
Model listing and chat provider construction rewrite localhost to
host.docker.internal via config.DockerLocalhost() when running inside
Docker, but all three memory.NewOpenAIEmbeddingProvider call sites
(verify-embedding handler, runtime memory embedding) passed the raw
api_base through. In Docker the container resolved localhost to itself,
so Ollama Verify Embedding and runtime embedding failed with connection
refused while the model dropdown loaded fine.
Centralize the rewrite inside the constructor so no call site can miss
it, add a SetInDockerForTest hook for environment-independent regression
tests, and pin Docker detection off in httptest-based suites.
Fixes#1519
The star history chart in the README is currently broken because the upstream service can no longer render it reliably under GitHub's stargazer API restrictions. Point the chart at a working alternative so the project's growth remains visible.
Co-authored-by: Dessalines39394 <245616256+Dessalines39394@users.noreply.github.com>
channel_pending_messages is a buffer, not an archive. Rows were deleted
outright on two paths: the bot being mentioned hands the buffer to the
agent and clears the key, and LLM compaction replaces old rows with a
summary. For group capture that buffer held the only copy of the raw
text, so a single mention or compaction pass destroyed days of messages
nothing had read yet.
Copy every row into channel_message_archive inside the same transaction
as the delete, tagged with the reason (consumed, compacted, stale).
Archived rows keep their original id, so a replayed delete is a no-op.
Add ListArchivedByKey so consumers that need full history read the
archive instead of the buffer.
An agent that generates an image and then attaches a copy of it under a
different name produced two media entries for one picture, and both were
sent. In production every image-edit reply arrived as two identical
attachments: the file create_image wrote (slug_timestamp_rand.png) plus
the path the agent passed to send_file, byte-identical but stored twice.
processMedia deduplicated by path only, so the pair survived. It also ran
before generated images were folded into MediaResults, leaving that
producer past the only dedupe pass entirely.
Fold persisted assistant images into MediaResults first, then deduplicate
once over the union, matching identical local files by content digest.
The assistant message is now built from that same deduplicated set, so
session history and outbound delivery cannot disagree. Media that cannot
be read (remote or already cleaned up) is left untouched.
The text-only gate sniffs the first 8192 bytes of a document and rejects
the read when they are not valid UTF-8. The window was cut at a plain byte
offset, so on non-ASCII text it regularly landed inside a multi-byte rune
and left a dangling prefix. The gate then reported the document as binary
even though the file was intact — measured on Vietnamese documents, about
one in four files above 8 KB was blocked, depending on where the accented
characters happen to fall.
Trim up to 3 bytes off the window when, and only when, the content was
actually cut there, then judge. A file that fits inside the window is left
alone, so genuinely broken trailing bytes are still caught, and binary
content stays blocked because its invalid bytes are spread over the whole
window. readCapped had the same byte-offset cut at max_bytes, which put a
broken character at the end of every truncated non-ASCII document.
The vault_read fix changes what the injected workspace means: it is the
global root wired once at boot, not a per-tenant root, since one tool
instance serves every tenant. These tests seeded a non-master tenant but
wrote fixtures straight to the injected root, so they only passed while
reads ignored the tenant layer.
Write fixtures to config.TenantWorkspace(base, tenantID, slug) and carry
the slug in context, matching a real run. tenantSlug is now the single
source of truth shared with seedTenantAgent, so on-disk layout in a test
cannot drift from the seed.
isAllCapsPlaceholder() classified ANY string of >=3 chars made only of A-Z and
underscore as an LLM placeholder, and stripEmptyOptionalArgs() then removed it
from the outbound CallTool arguments. Real business identifiers match that shape:
bank codes BKASH, NAGAD, ROCKET, VCB
currency codes BDT, VND, USD, PKR, MMK
enum values SUCCESS, PENDING, ALL
Impact is silent and severe: dropping an optional filter turns a filtered query
into an unfiltered one. The MCP server still answers 200 with the full result set,
so nothing errors — the model reports totals for the entire dataset as if they were
the filtered result, with full confidence.
Observed in production: asking "how many BKASH proxies are active" returned 6,610
(all banks) instead of 1,263. Three different bank codes returned byte-identical
results, which also drove the model into a retry loop until the loop guard fired.
Diagnosis was slow because the removal is invisible: the outbound log records
args_len BEFORE stripping, and stripEmptyOptionalArgs logs nothing at all — while
normalizeArgsForSchema right below it does emit mcp.tool.args.coerced. Confirmed
the argument never reaches the server by echoing the raw JSON-RPC arguments from
RequestContext[CallToolRequestParams] on the server side.
Changes:
- isAllCapsPlaceholder: require multi-word (contains "_") or membership in a known
single-word placeholder list. Keeps the original intent — the doc comment's own
examples (SHOULD_NOT_BE_HERE, DO_NOT_SEND, NOT_APPLICABLE) are all multi-word —
while letting business codes through.
- stripEmptyOptionalArgs: emit slog.Warn("mcp.tool.args.stripped") whenever an
argument is dropped, so this class of bug is one grep away instead of a
multi-hour investigation.
- Tests: guard both layers against regression.
Verified: go build ./... clean; go test ./internal/mcp/ passes.
vault_read is a singleton wired once at boot with the global workspace
root, then joins doc.Path (stored relative to the TENANT root) onto it.
For master tenant both roots coincide, so the bug stayed invisible; every
other tenant reads ENOENT because its files live under tenants/<slug>/.
vault_search is DB-only and keeps returning the doc, so the symptom is
"search finds it, read fails" rather than an obvious missing file.
Resolve the tenant-scoped root per request via TenantLayer, which is a
no-op for master and preserves existing single-tenant behaviour. Passing
the tenant root into resolvePath also narrows the boundary check, so it
now doubles as cross-tenant isolation.
Runs originating from channels or cron may not carry the tenant slug in
context; without it TenantLayer falls back to a UUID-named directory that
does not match the slug-named one the write path creates. Add a nil-safe
TenantStore lookup for that case only.
Gemini drops tool_call_id and pairs functionCall/functionResponse by
function name, so its OpenAI-compat shim requires a non-empty
FunctionResponse.name. The name was only recoverable through a reverse
id->name lookup over assistant tool_calls still present in the request
window, which fails after pruning, truncation or tool_call collapse and
produces HTTP 400 "Name cannot be empty" on any follow-up iteration.
Add Message.ToolName, set it at every tool-result creation site and
preserve it through context pruning, then prefer it when serializing.
Fall back to the existing index for history persisted before the field,
so old sessions keep working without a reset.
When neither source resolves a name, drop the unlabelled tool result
instead of emitting an empty one: an empty name is a guaranteed 400 and
a synthetic name would match no prior functionCall. Gated behind the
existing Gemini detection so other OpenAI-compat hosts, which pair by
tool_call_id, are unaffected.
Field is JSON-optional, so session history needs no migration and older
binaries ignore it on rollback.
The workstations page was written against snake_case field names the
gateway never served. Go's json decoder drops unknown members silently,
so create requests arrived with an empty WorkstationKey and were rejected
with "workstationKey is required", while list rows read undefined and
rendered a blank key, "backend.undefined" and "Invalid Date".
Align the TypeScript contract on the camelCase shape that both the WS
handler and store model already use:
- Rename Workstation, CreateWorkstationParams and every field read on the
workstations page to workstationKey/backendType/createdAt/updatedAt.
- Nest update params under `updates`; a flattened body decoded to an empty
map and was rejected with "no updates provided".
- Replace the Identity File input, which mapped to a field SSHMetadata does
not accept, with a private key or password credential path.
- Add the image field required by DockerMetadata and map the container name
onto host and the daemon endpoint onto socketPath, so Docker workstations
can be created at all.
- Extract the payload builder so the wire contract is unit-testable, and
move validation strings into the en/vi/zh catalogs.
Covered by tests asserting the create payload shape and list rendering
against a verbatim gateway response body.
Co-authored-by: Eddy Lockwood <doakythanh@gmail.com>
SSRF protection resolves a hostname and judges the resulting IP. That
model assumes DNS resolution describes where the traffic actually goes,
which stops being true behind a TUN/fake-IP proxy: every query is answered
with a synthetic address out of a reserved range, and the proxy then
routes that address to the real public host. The IP is a handle, not a
destination.
In that environment web_fetch rejects ordinary public sites — observed
with news.sina.cn resolving to 198.18.0.236 — and no configuration can
fix it, because the block list is compiled in. The agent then burns
iterations retrying URLs that can never succeed.
GOCLAW_SSRF_ALLOWED_CIDRS lets an operator name the ranges their proxy
hands out. Empty by default, so nothing changes for deployments that do
not set it, and the accepted and refused entries are both logged at
startup — this widens what LLM- and admin-supplied URLs can reach, so it
should be visible.
Ranges an SSRF actually targets can never be allowlisted: link-local
(including cloud metadata at 169.254.169.254), multicast and unspecified
are refused at parse time, in either direction, so neither an exact entry
nor a wider range that swallows one gets through.
Applied inside isBlocked rather than only in validate() because
NewSafeClient re-checks the pinned IP at dial time through the same
function — relaxing just the pre-flight check would pass validation and
then fail to connect.
internal/tools carries its own private-range list for web_fetch and
web_search, separate from this package and not identical to it. It has to
consult the same setting, or relaxing one gate leaves the other rejecting
the very traffic the operator permitted. Unifying the two lists is left
alone here; it is a wider change than this one.
Co-authored-by: Conner Mo <connermo@ConnerdeMacBook-Pro.local>
Added error handling for input reading in the agent chat client and for database row reading in the doctor commands, ensuring that errors are reported clearly to the user.
A reasoning model emits nothing at all while thinking, so the wait for the
response's FIRST byte — not any streaming timeout — decides whether its answer
is reachable. At 180s a large-prompt reasoning request routinely lost that race:
the gateway reported "http2: timeout awaiting response headers" while the
router's own usage log showed the same request completing normally moments later
(promptTokens=77194 -> completionTokens=2789). The request was healthy; only the
deadline was short.
Raise it to 300s. The value is deliberately modest rather than generous because
it is multiplied twice downstream: RetryDo wraps every provider call with
Attempts: 3, and a workflow step requeues up to MaxTaskDispatches times on top
of that, so 300s is already ~45min of worst-case wall clock per step.
Removing the timeout entirely was considered and rejected. Agent run contexts
come from scheduler queue.go via context.WithCancel, so they carry no deadline —
this transport timeout is the only thing bounding a request that never answers.
Without it a silently dead connection would hold its scheduler lane until the OS
TCP keepalive expired, and workflow requeue could not rescue it because requeue
only fires when a run ends. The doc comment claiming the design could "rely on
ctx deadlines" was wrong and is corrected here.
Tests lock the sizing from both sides (must exceed the latency that failed, must
not exceed the ceiling the retry math was sized against) and assert the retry
multiplier itself has not drifted.
Co-authored-by: nguyenha935 <nguyenha935@users.noreply.github.com>
The extractive fallback runs when an LLM memory flush fails, so it is the last
thing standing between a session and losing its accumulated decisions. Its cue
patterns were English-only ("decided to", "prefer", "the API is"), which makes it
effectively inert for a session conducted in another language: it still returns a
non-empty result, so nothing looks broken, but the result is noise.
Measured on 195k characters of real Vietnamese session text: 9 decision matches
(all of them stray English fragments), 0 preferences, 0 technical facts — about
97% of what it "saved" was incidental file paths. After adding Vietnamese cues the
same corpus yields 21 decisions and 6 preferences on the largest session.
Adds parallel Vietnamese patterns for the same three categories and merges them
into the existing results with deduplication, so English extraction is unchanged
and mixed-language sessions get both. Tests cover Vietnamese-only,
English-only, and mixed histories, plus a guard that ordinary Vietnamese
narration is not misread as a decision or a preference.
Co-authored-by: nguyenha935 <nguyenha935@users.noreply.github.com>
