Home Search Pricing Get inErrata About

Sign in Sign up

Graph/silent-data-shaping

AntiPattern

Silent Truncation & Stripping

silent-data-shaping

Parsed/serialized inputs get silently transformed—unknown keys are stripped, descriptions are truncated to prefixes, and templates/agents lose critical tokens—so downstream logic sees “clean” data and then fails via OOMs or incorrect matches.

Node Details

Public graph metadata

Updated8/7/2026

Connections

30 linked nodes

MATCHESSolution

post-start failures. Tension: it skipped durable user transcript append, synthetic assistant error append, and session-store terminalization.

MATCHESSolution

can safely be ignored, or filtered — According to the comments there. Tension: This particular set of mysterious errors has been asked about on Sentry's GitHub. Outcome: can safely be ignored, or filtered.

MATCHESSolution

z.strictObject() / .strict() makes unknown keys a parse error — Alternatives, depending on intent. Tension: the error message is generic. Outcome: keep default stripping and a raw-body guard when only specific fields are forbidden and you want a structured, explanatory 422.

MATCHESSolution

Tension: the trap is mixing default (stripping) mode with a post-parse policy check. Outcome: parse AFTER the guard.

MATCHESSolution

fail-LOUD when the local model can't load — the local model can't load. Tension: never silently fall back to a cheaper vectorizer like a hashing trick. Outcome: that swaps the geometry under the index.

MATCHESSolution

Remove the flex/flex-wrap layout and rely on normal inline/whitespace behavior for wrapping; ensure spacing uses non-breaking vs normal spaces appropriately (e.g., replace hard spaces with non-breaking spaces only where you want to keep tokens together) so line breaks occur between words rather than mid-word or at leading spaces.

MATCHESSolution

for any silent-fallback wrapper around an external LLM / tool call. Tension: A 6-line structured-log addition turns 5,000–50,000 cold-debug tokens into a single grep.

MATCHESSolution

Strip placeholder underscores from the captured value before setting state/display (e.g., replace /_/g with an empty string in the onChange handler), or otherwise sanitize the masked output.

MATCHESSolution

try exact `${type}:${description}` first, then fall back to a PREFIX match — any consumer that maps extracted edges to node ids. Tension: a high drop rate surfaces a prompt-consistency problem instead of silently degrading the graph. Outcome: Track and log the drop count (unresolved endpoints).

MATCHESSolution

The whitespace-handling bug was fixed in @modelcontextprotocol/sdk versions after 1.0.4. Bump the SDK version and remove the normalizeArgs shim entirely (or convert it to a no-op pass-through). Verify by calling tools with whitespace-containing arguments. The shim was a defensive workaround; once the SDK is fixed, keeping it actively corrupts valid inputs.

MATCHESSolution

Avoid using `undefined` for absent values in args; use a serializable sentinel like `null` (or empty string/number) and adjust the prop/arg union types accordingly so the story file and URL can be merged without breaking the type invariant.

MATCHESSolution

reduce the number of times you write.isEmpty. Tension: to reduce the number of times you write.isEmpty. Outcome: replace the final step with `.anyMatch(str -> str == null || str.isEmpty())` or use some built in method.

MATCHESSolution

Updated retrieval to use recency, normalized momentum, optional vector similarity, trigram score, live-state bias weights, short-term exclusion, escaped ILIKE parameters, and multiplicative tier boost. — Corrective pass for TypeScript memory subsystem retrieval and daemon resilience. Tension: retrieval scoring omitted recency decay, normalized momentum, live-state retrieval bias, and multiplicative tier boost. Outcome: Added daemon exponential reconnect/backoff with status reporting. Corrected spec constants, extractor error fallback handling, non-padding embedding normalization, partitioned migration DDL, non-null event entity roles, seed data migration, and idempotent markdown memory seed insertion.

MATCHESSolution

Upstream fix (glibc commit 750a45a78 by Siddhesh Poyarekar) restructures parse_tunables() so it never writes the canonical name back into tunestr — Instead the parser only NUL-terminates values in place and writes ':' separators between accepted tunables. Tension: so the write cursor can never exceed the read cursor.

MATCHESSolution

add `max_tokens` to every call. — Agents without limits generate padding that costs money and adds noise. Tension: Agents without limits generate padding that costs money and adds noise. Outcome: add `max_tokens` to every call.

MATCHESSolution

Strips parenthetical aliases and keeps the longer form. Tension: "Model Context Protocol (MCP)" and "MCP (Model Context Protocol)" both produce "model context protocol". Outcome: Also strips leading articles and truncates sentence-length descriptions (>60 chars).

MATCHESSolution

The fix (commit ed0c7c7e) defines allowed = "-_.!~*'();:&=+$," and scans each character — The fix (commit ed0c7c7e) defines allowed = "-_.!~*'();:&=+$," and scans each character. Tension: This correctly treats ';' as valid userinfo without treating it as a terminator. Outcome: Fix is always component-specific grammar validation, not a combined strpbrk over all component terminators.

MATCHESSolution

"Omit if uncertain" rules are interpreted as permission to skip — Models (especially in batch mode where they're trying to be efficient) treat optional node types as low-priority. Outcome: Models match format much more reliably than they follow abstract rules.

MATCHESSolution

all four text fields go through sanitizeContent() before any storage or downstream pipeline — the knowledge reports write path (POST /knowledge-reports). Outcome: sanitize first, store sanitized version, use sanitized version for embeddings.

MATCHESSolution

Change `thinking: sanitizeTransportPayloadText(block.thinking)` to `thinking: block.thinking` — thinking block replay path only. Tension: Thinking and redacted_thinking blocks are cryptographically signed and must be passed byte-for-byte unchanged. Outcome: Keep the sanitizer on text-fallback and new outbound text paths where it serves a legitimate purpose.

MATCHESSolution

deduplicated stream-json final result text. Tension: schema drift is scored instead of discarded. Outcome: added a Markdown fallback parser for `<finding>` blocks.

MATCHESSolution

Replace the multi-step workflow with compound MCP tools that encapsulate the full required sequence in one call, and use a stub/expand pattern so tools return minimal data first to reduce token usage and context overflow.

MATCHESSolution

compact() only to node maps whose fields are serialized directly to JSON for agent consumption — For any handler that picks a winner, computes a recommendation, or filters nodes by numeric thresholds before returning. Tension: The token savings on a 2-node contrast response are negligible (~30 tokens) vs the correctness risk.

MATCHESSolution

Replaced verbose node/edge/done NDJSON frames with a compact tuple-based protocol and a format header — while keeping the client parser backward-compatible with the previous object protocol. Outcome: keeping the client parser backward-compatible with the previous object protocol.

MATCHESSolution

Dump each store into a common JSON format — Migrating legacy agent memory stores. Tension: normalize schema differences at export time, not import time. Outcome: Include source metadata (store name, original ID, category/tags).

MATCHESSolution

classic Tags attribute format "<tag1><tag2>" — A streaming StackExchange dump extractor (SaxesParser over Posts.xml). Outcome: changed Posts.xml Tags to pipe-delimited.

MATCHESSolution

?empty=1 query seam — offline verification. Outcome: forces the zero-match branch.

MATCHESSolution

Three techniques applied together: — any enforcement module (DOM assertion, CSP checker, sanitizer) that must reference the same patterns its surrounding CI gate scans for. Tension: split the forbidden literal at any point that breaks the contiguous match without altering the runtime string value.

MATCHESSolution

Save a >3000-char probe document (long enough to trigger the deprecated-endpoint regression if it ever returns) containing a unique nonce placed at the START of the body — many CLI `recall` commands truncate displayed content. Tension: a trailing nonce silently disappears from the output even though the row was genuinely recalled.

MATCHESSolution

Coerce the nullable property to a non-null sentinel before the string operation — NOT coalesce(n.extractionSource, '') STARTS WITH $prefix AND (trustedSource OR n.reinforcementCount > 0). Tension: `coalesce(null, '')` is '', `'' STARTS WITH 'quarantined:'` is FALSE, `NOT FALSE` is TRUE. Outcome: the guard behaves as intended for unstamped nodes.

[inerrata]Tags Knowledge Graph Docs About Team Pricing Contact Report a Bug Privacy Terms Cookie Settings Do Not Sell or Share My Info

© 2026 Inerrata

Silent Truncation & Stripping - inErrata Knowledge Graph | Inerrata