[sergo] Sergo Report: New 29th Linter ssljson — Impl Narrower Than Its Own Embedded Spec - 2026-06-16

[github-actions[bot]]

Executive summary

The custom-linter registry grew 28 → 29: ssljson is the new addition — and unlike its 28 siblings it is not a Go-pattern linter at all, but a validator of .github/skills/*/ssl.json skill artifacts against the SSL Pass-4 enum + graph rules. It has never been audited in 38 runs. The audit found the implementation is narrower than the linter's own embedded spec: the repo's own ssl.json enumerates the Pass-4 rules, and two of them are unimplemented. Filed 2 spec-backed issues (sg38a1, sg38a2). Reconciliation also recovered an uncommitted run and confirmed two prior landings.

Tool & registry updates

Surface	R36 (06-14)	R38 (06-16)
Serena LSP tools	23	23 (stable R4–38)
Custom linters (main.go)	28	29 (+ssljson)
doc.go All N	"All 25" (3-way drift)	"All 29" — synced
CI enforced flags	13	14 (+-timeafterleak)

ssljson (pkg/linters/ssljson/ssljson.go) uses the anchor-package-once idiom (run returns early unless pass.Pkg.Path()==anchorPkg, :207) so it emits diagnostics exactly once per golint-custom invocation — clean.

Strategy — 50/50 split

• Cached (proven methodology): new_linter_precision_audit + enforce_readiness_run_on_prod + phantom_issue_reconcile. Reconciled all 9 open sergo issues + issues created since last_run via gh api.
• New (exploration): first-ever audit of ssljson, a project-specific JSON-artifact validator — a linter class never examined in 38 runs. New land-argument crystallized: impl_narrower_than_own_spec (diff the implementation against the spec the artifact itself embeds).
• Run targets: ≥2 distinct spec-backed findings, ≤3 non-duplicate issues, full reconciliation. All met.

Findings

The repo's own .github/skills/ssl/ssl.json:150 (step_verify_validate description) defines the Pass-4 rule set the analyzer is meant to enforce:

"JSON syntax, required fields, enum membership, unique IDs, entry pointers, transition targets, graph integrity"

ValidateDoc (ssljson.go:93-149) implements enum membership + entry/transition pointers, but two named rules are missing:

sg38a1 — graph-integrity FN: logic-step scene_id parsed but never validated

• SSLStep.SceneID is decoded (ssljson.go:61) and populated on all 19 logic steps across the 4 production ssl.json files (reporting, jqschema, error-messages, ssl), yet ValidateDoc never references it.
• Every other cross-reference is checked (entry_scene→scenes, entry_logic_step→steps, transition target→scenes, step next→steps). The step→scene back-reference is the only declared relationship left unchecked — a parsed-but-dead field.
• A step whose scene_id names no scene (typo / rename orphan) passes with zero diagnostics.
• Fix: add a Rule 8 in the per-step loop mirroring the entry_logic_step check shape (:116-118). Small, single-file + one testdata case.

sg38a2 — duplicate scene/step IDs silently deduped: the spec's "unique IDs" rule is unimplemented

• sceneIDs/stepIDs are built as map[string]bool (ssljson.go:96-103), so two scenes (or steps) sharing an id collapse silently — the second overwrites the first and downstream reference checks resolve against an ambiguous table.
• The spec lists "unique IDs" as an explicit Pass-4 rule; there is no uniqueness check anywhere in ValidateDoc.
• Latent today (the 4 prod files have unique IDs), exactly the class to close before CI enforcement.
• Fix: detect collision while inserting (if m[k] { report }; m[k]=true) in each of the two existing loops. Small.

Reconciliation (since last committed run, 06-14)

Recovered run + landings + open-issue status

• Recovered R37 (06-15, memory never committed): run 27525869216 audited timesleepnocontext (26th) → timesleepnocontext precision: false positive on time.Sleep inside deferred cleanup closures #39323 precision (FP on time.Sleep in deferred cleanup closure, mcp_inspect.go:138) + timesleepnocontext enforce-readiness: zero production violations after the defer-cleanup FP fix — add to CI #39324 enforce-readiness. Both OPEN. Reconstructed into strategy history. Not refiled — not mine.
• timeafterleak (28th linter): 3 prod loop+select time.After sites leak a timer per iteration — convert to time.NewTimer/Stop, the [Content truncated due to length] #39180 (timeafterleak enforce) LANDED in code: -timeafterleak is now in cgo.yml:1123 (14 flags); the 3-site conversion shipped, though the issue is not auto-closed.
• doc-sync: linters/doc.go still says "All 24 active analyzers" after two new linters (hardcodedfilepath, timesleepnocontext) land [Content truncated due to length] #38787 (doc-sync) resolved-in-code: doc.go now reads "All 29 active analyzers" with all bullets present; issue still open but no code gap remains.
• Stale: execcommandwithoutcontext precision: false positive on nil-guarded exec.Command fallback — autofix injects a nil context that pa [Content truncated due to length] #38281/execcommandwithoutcontext enforce-readiness: propagate context in connectStdioMCPServer (2 sites), add nolint support, then enfo [Content truncated due to length] #38282 (execcommand) now 6 days — complex control-flow, do-not-refile. hardcodedfilepath precision: hasFormatVerb misses flag/width/precision verbs (%05d, %-10s, %8.2f) and 9 verb letters (%x %t %c % [Content truncated due to length] #38788/ctxbackground precision: isContextBackgroundCall matches identifier name "context" syntactically while the same file detects the [Content truncated due to length] #38789 open 4d. timeafterleak precision: a stored time.After channel (timer := time.After(d); <-timer) inside a loop+select escapes detection — [Content truncated due to length] #39181 (timeafterleak stored-channel FN) open.

Metrics

Metric	Value
Run	R38 (27596186434)
Findings	6 (2 filed + 4 reconcile/state)
Issues created	2 (sg38a1, sg38a2) — 0 duplicates
Success score	9 / 10
Cumulative	38 runs · 281 findings · 76 tasks · avg 8.78

Recommendations & next-run focus

• R39: confirm sg38a1/a2 posted (ssljson title-regex) and watch for landing. ssljson has more graph rules worth auditing: empty required ids, entry_scene == "" slipping past the != "" guard (:106), and unreachable-scene / terminal-reachability checks. Also assess -ssljson enforce-readiness (the 4 prod files validate clean under current rules).
• Watch for a 30th linter (main.go registry delta). Reverify execcommandwithoutcontext precision: false positive on nil-guarded exec.Command fallback — autofix injects a nil context that pa [Content truncated due to length] #38281/execcommandwithoutcontext enforce-readiness: propagate context in connectStdioMCPServer (2 sites), add nolint support, then enfo [Content truncated due to length] #38282 — if 7+ days dead, switch to noop/report.
• Process: the 06-15 uncommitted-memory gap reinforces always querying gh api for sergo issues created since last_run before picking a strategy.

References:

• §27596186434 (this run)
• §27525869216 (recovered R37, 06-15)

Generated by 🤖 Sergo - Serena Go Expert · 284.3 AIC · ⌖ 13.8 AIC · ⊞ 5.3K · ◷

• expires on Jun 16, 2026, 9:39 PM UTC-08:00

[github-actions[bot]]

This discussion was automatically closed because it expired on 2026-06-17T05:39:55.639Z.

Closed by Workflow