speckit-aee-route-evidence
Skill4 starsChanged 55 days agoMITView on GitHub
Route oversized logs and sources into raw evidence slices for AEE gates
- Reads credentials
1 other project uses a version of this.
Adopted by
Most recently taken up by
- electrohire/spec-kit-evaluatorword for word, first seen Oct 25, 2026
Other versions1 project uses a version of this.
Word for word · 1
- electrohire/spec-kit-evaluator★ 3100% same
Edited · 0
None found.
Wrote their own version · 0
None found.
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository25 other files it gives its agents.
- speckit-aee-assessSkill
- speckit-aee-challengeSkill
- speckit-aee-draft-claimsSkill
- speckit-aee-gapsSkill
- speckit-aee-gateSkill
- speckit-aee-message-securitySkill
- speckit-aee-report-savingsSkill
- speckit-aee-reviewSkill
- speckit-aee-traceSkill
- speckit-aee-verifySkill
- speckit-analyzeSkill
- speckit-checklistSkill
- speckit-clarifySkill
- speckit-constitutionSkill
- speckit-convergeSkill
- speckit-evaluator-composeSkill
- speckit-evaluator-reportSkill
- speckit-evaluator-routeSkill
- speckit-evaluator-runSkill
- speckit-implementSkill
- speckit-planSkill
- speckit-specifySkill
- speckit-tasksSkill
- speckit-taskstoissuesSkill
- spec-kit-aeeSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 4 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 34
Never route secrets: any path component (directories included, not just the file name) suggesting credentials, private keys, service …
- Distinctive: line 43
Run the policy helper, which selects the backend and enforces the evidence rule:
- Very specific: line 60
When the helper answers read_directly, read the file directly.
- Distinctive: line 79
Never assume omitted lines are irrelevant.
- Distinctive: line 88
Keep mandatory always-on root instructions short.
- Distinctive: line 90
Do not describe routed evidence as proof, certification, or truth.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Aee Route Evidence Skill
AEE Route Evidence
Route a file that is too large to load directly into verbatim raw slices, so
speckit.aee.assess and speckit.aee.challenge gates reason over evidence
instead of whole files.
User input
$ARGUMENTSAccept file=<path>, mode=<error_log|heavy_code|agent_context>, and
query="<text>". Infer the mode from the file kind when omitted: logs, stack
traces, and CI output are error_log; source files are heavy_code; long
instruction references such as AGENTS.md or agent-context/*.md are
agent_context.
Prerequisites
- The target file must be inside the project root. Never follow symlinks or read outside the project root.
- Never route secrets: any path component (directories included,
not just the file name) suggesting credentials, private keys,
service accounts, or
.envcontent is refused fail-closed, with_/-spellings treated alike. The guard is name-based: it cannot detect secret content hiding under an innocuous name, so do not route a file whose contents you have not accounted for.
Execution
Run the policy helper, which selects the backend and enforces the evidence rule:
python .specify/extensions/aee/scripts/python/aee_token_economy.py route \
--file "<path>" --mode "<mode>" --query "<query>"Backend selection, in order:
- The upstream
token-routerscript plus Ollama, when both are available. The local model is a line router, never a summarizer: it returns line ranges, and the helper extracts the exact raw slices. - Deterministic fallback (pure-Python literal match windows — no external search tool is invoked) when token-router or Ollama is absent. State the backend in the notes.
When the helper answers read_directly, read the file directly; routing a
small file wastes more than it saves. When it answers no_matches, sharpen
the query or widen the window instead of inventing ranges.
Trust and limits: the token-router is executed code, so the helper only
runs a router found inside the project root or its parent — an
out-of-tree TOKEN_ROUTER_HOME is ignored and routing falls back
deterministically. When the router backend runs, the record names the
router path and its SHA-256 (router field). Files over the byte cap
(256 MB by default, --max-file-bytes) are refused with a slicing hint
rather than loaded whole.
AEE evidence rule
- The router selects lines; it never analyzes. Its prose is
unsupportedunless grounded in a verbatim slice. - Evidence citations are
path:start-endranges over the original file. - Raw slices keep their AEE evidence
kind(observedfor tool output and logs,artifactsource quality where the vocabulary applies). - Never assume omitted lines are irrelevant. If a slice looks too narrow, expand the range and re-read before making claims.
- Save evidence notes under
.specify/extensions/aee/evidence/with file paths, line ranges, the backend used, and timestamps.
Guardrails
- Assess and challenge gates must see the raw slices, never a compressed or summarized substitute. Compression is for transport, not for evidence.
- Keep mandatory always-on root instructions short; routing cannot save tokens after a long root instruction file has already been injected.
- Do not describe routed evidence as proof, certification, or truth.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

