Ready to supercharge your AI agents?

Install any skill in seconds. Free to start, no credit card required.

Observal Advanced Operations · bilal140202 | DecimalAI

bilal140202/observal-advanced

SkillScore100Maintained

Advanced Observal operations including session reconciliation, CLI upgrades and downgrades, complete uninstallation, and local fallback mode for offline use. Use when the user wants to reconcile sessions, upgrade or downgrade the CLI, uninstall Observal, or write agent configs locally when the server is unreachable.

agent-toolingby bilal140202

Links this skill to your agents and keeps it current with the author’s updates.

Automatic

DevOps & SRE pack

1 · One-time — install the CLI (docs ↗)

pip install decimalai

2 · Get the skill — writes .claude/skills/bilal140202-observal-advanced/SKILL.md

decimalai skills pull bilal140202-observal-advanced --out .claude/skills/

No CLI? Get the same file — curl · raw file

curl -fsSL --create-dirs -o .claude/skills/bilal140202-observal-advanced/SKILL.md https://app.decimal.ai/s/bilal140202-observal-advanced/SKILL.md

Raw file: /s/bilal140202-observal-advanced/SKILL.md ↗ · machine-readable summary at /s/bilal140202-observal-advanced.json ↗

In productionGlobal

how this skill behaves on real traffic · last 30 days

No production data for this skill yet — usage numbers appear here once agents start running it.

Anonymized aggregate across every org using this skill — proof from real production use, not a demo.

Does this skill help? · Benchmark: with vs. withoutsign in to run this on your agents

22 test cases from the publisher’s eval.yaml · each run twice on gemini-3.6-flash — with the skill injected, then without · last run 51d ago

Yes — +86% lift · at higher cost+49% tokens0% turns100% of whole cases pass with skill (22/22)

Test case	Without → With	Effect	Δ tokens	Δ turns
case-01	✗→✓	▲ Improved	1%	0%
case-02	✗→✓	▲ Improved	-3%	0%
case-03	✗→✓	▲ Improved	13%	0%
case-04	✗→✓	▲ Improved	52%	0%
case-05	✗→✓	▲ Improved	27%	0%

Showing 5 of 22 cases

SkillSafety

✓ 3 of 3 passed

Code scan15 of 15 checks clean

0 critical · 0 warning · 0 info — scanner v4 — 8/2/2026

live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing

IntentSafe

LLM judge · does the skill do what it claims

ContentClean

LLM judge · policy & harmful-content review

Source

Imported

bilal140202 / the-lord-of-the-skills

skills/gondor/claude-code/BlazeUp-AI__Observal/SKILL.md

View on GitHub ↗

License MITImported 8/2/2026Version v1

Skill contents

SKILL.md

Observal Advanced Operations

Procedure: Reconcile Sessions

Push local session JSONL to server for full trace context. Normally automatic via hooks.

Manual run for crash recovery:

bash
python -m observal_cli.cmd_reconcile

Not a Typer command. No flags needed.

Procedure: Self-Manage CLI

bash
observal self status
observal self upgrade
observal self upgrade --version 2.5.0
observal self downgrade --list
observal self downgrade --version 2.4.0
observal self rollback

Procedure: Uninstall

Completely removes Observal. Destructive and irreversible.

bash
observal self uninstall

Requires typing confirm (not --yes). Selective flags:

--keep-config: preserve ~/.observal/
--keep-cli: keep the CLI binary
--keep-repo: keep the cloned repo
--repo-dir PATH: specify repo location

Procedure: Local Fallback Mode

Use only when a command exits with Connection failed or Not configured.

| IDE | User-scope path | Project-scope path | |---|---|---| | Claude Code | ~/.claude/agents/<name>.md | .claude/agents/<name>.md | | Kiro | ~/.kiro/agents/<name>.json | .kiro/agents/<name>.json | | Cursor | ~/.cursor/rules/<name>.mdc | .cursor/rules/<name>.mdc | | Gemini CLI | ~/.gemini/agents/<name>.md | .gemini/agents/<name>.md | | VS Code | ~/.config/Code/User/agents/<name>.md | .vscode/agents/<name>.md | | Codex CLI | ~/.codex/agents/<name>.md | .codex/agents/<name>.md | | Copilot CLI | ~/.config/github-copilot/agents/<name>.md | .github/copilot/agents/<name>.md | | OpenCode | ~/.opencode/agents/<name>.md | .opencode/agents/<name>.md |

Kiro (~/.kiro/agents/<name>.json):

json
{"name":"<name>","description":"<desc>","prompt":"<prompt>","model":"claude-sonnet-4-20250514","mcpServers":{},"tools":["*"],"resources":["skill://~/.kiro/skills/*/SKILL.md"]}

Claude Code, Gemini CLI, VS Code, Codex CLI, Copilot CLI, OpenCode (markdown):

markdown
---
name: <name>
description: <desc>
---
<prompt>

Cursor (.mdc):

markdown
---
name: <name>
description: <desc>
---
<prompt>

After writing locally, remind the user to run observal agent create once the server is reachable.

Output Contract

One sentence stating intent.
The exact command in a fenced code block.
The result: success / specific error.
The next action, or "done".

Illustrative Want to see what bilal140202/observal-advanced would put in front of your model? Preview the rendered body and check its trigger phrases — no LLM call, no signup.

Latest benchmark · publisher runVerified

8/8/2026, 16:19 UTC · 22 cases

Lift

+86%

100% vs 14%

Agent turns

1.0 vs 1.0

Tokens

+49%

1,504.41 vs 1,012.18

fail → pass · 19pass → pass · 3

Per-case results (22)

Case	Status	Duration (ms)			Turns			Tokens			Tool calls
Case	Status	Without	With	Δ	Without	With	Δ	Without	With	Δ	Without	With	Δ
case-01 My shell crashed mid-process and I need to manually upload my local JSONL session logs to the server so trace context is complete. What is the execution step to reconcile my sessions? Structure the output by first stating your goal in one sentence, then giving the exact terminal command in a code fence, followed by the result status, and ending with the next action or 'done'.	fail→pass	5,723	1,666	-71%	1	1	0%	1,121	1,133	+1%	0	0	—
case-02 I need to remove Observal completely from my system, but I want to retain my saved preferences inside ~/.observal/. What command should I execute? Please respond with a single sentence stating intent, the precise command block, the resulting status, and the subsequent step or 'done'.	fail→pass	5,941	2,100	-65%	1	1	0%	1,196	1,156	-3%	0	0	—
case-03 When executing Observal commands, I ran into a 'Connection failed' error. I want to set up a local fallback agent named 'analyzer' for Claude Code at the project level. How do I construct this fallback? Please provide your response structured as: a one-sentence intent summary, the fenced command/script block, the execution result, and the next action to perform once the server is back online.	fail→pass	6,645	3,566	-46%	1	1	0%	1,334	1,513	+13%	0	0	—
case-04 I need to upgrade the Observal CLI on my machine specifically to release version 2.5.0. How do I perform this version upgrade? Please state your intent in one sentence, present the command in a fenced block, note the result status, and end with the next action or 'done'.	fail→pass	3,538	2,207	-38%	1	1	0%	802	1,222	+52%	0	0	—
case-05 I am troubleshooting a CLI issue with Observal and want to see all available earlier versions that I can downgrade to. What CLI command displays this list? Provide your answer formatted as a one-sentence goal statement, the command block, the expected outcome, and the next action or 'done'.	fail→pass	5,332	2,823	-47%	1	1	0%	1,010	1,283	+27%	0	0	—
case-06 I need to downgrade my local Observal CLI installation to version 2.4.0 after a regression in the latest release. What is the CLI invocation? Format your response with a single-sentence intent, the fenced command block, result description, and final step or 'done'.	fail→pass	5,728	1,901	-67%	1	1	0%	1,051	1,100	+5%	0	0	—
case-07 After upgrading Observal, the CLI is throwing errors, and I need to immediately revert to the previously installed CLI binary version. What command performs this action? Structure your output as a one-sentence intent summary, code block, result status, and next step or 'done'.	fail→pass	3,948	2,622	-34%	1	1	0%	777	1,311	+69%	0	0	—
case-08 I need to remove Observal components from my system while retaining only the CLI executable binary itself. Which command flag combination achieves this? Structure the answer as a one-sentence intent statement, code block, result outcome, and next action or 'done'.	fail→pass	5,883	3,305	-44%	1	1	0%	1,152	1,435	+25%	0	0	—
case-09 I am uninstalling Observal from my machine but want to make sure the cloned repository directory remains untouched on disk. What uninstall command should I execute? Structure your response as a single sentence stating intent, the command block, result summary, and next step or 'done'.	fail→pass	3,752	3,058	-18%	1	1	0%	688	1,393	+102%	0	0	—
case-10 I want to uninstall Observal, specifying that the local repository source is located at `/opt/src/observal`. What command should be run? Format your output with a one-sentence intent summary, fenced command block, result, and next action or 'done'.	fail→pass	4,177	1,818	-56%	1	1	0%	910	1,150	+26%	0	0	—
case-11 Observal CLI returned 'Not configured' while offline. I need to create a user-scope local fallback agent named 'helper' for Kiro. What file path and contents should be written? Format the response with a single sentence stating intent, a code block creating or writing the file, result outcome, and next action.	fail→pass	4,931	4,038	-18%	1	1	0%	1,117	1,753	+57%	0	0	—
case-12 While working offline with Kiro, Observal output 'Connection failed'. I want to set up a project-level local agent fallback named 'patcher'. How do I construct this fallback? Please provide your answer with a one-sentence goal statement, the code block, expected result, and next action once online.	fail→pass	5,852	4,320	-26%	1	1	0%	1,139	1,743	+53%	0	0	—
case-13 Observal CLI returned 'Connection failed' during a network outage. I need to set up a user-scoped fallback agent named 'architect' for Cursor. Where should this file be saved and what format should it use? Structure your answer as a one-sentence goal, a command/code block, result summary, and next action.	fail→pass	5,476	3,600	-34%	1	1	0%	1,209	1,655	+37%	0	0	—
case-14 When running Observal, I received a 'Not configured' response. I need to define a project-scope fallback agent named 'linter' for Cursor. What path and format are required? Format your response with a single-sentence intent summary, code block, result status, and next step.	fail→pass	5,308	3,179	-40%	1	1	0%	1,201	1,488	+24%	0	0	—
case-15 Observal exited with 'Connection failed'. I need to configure a project-level fallback agent called 'reviewer' in VS Code. What file location and frontmatter format should be created? Structure your response as a one-sentence intent statement, code block, result status, and next action.	pass→pass	5,081	4,143	-18%	1	1	0%	1,051	1,642	+56%	0	0	—
case-16 Observal commands fail with 'Connection failed'. I want to create a project-level fallback agent named 'scout' for Gemini CLI. What path should be used? Provide your response structured with a one-sentence intent summary, the code fence block, execution result, and next step.	fail→pass	4,327	3,710	-14%	1	1	0%	903	1,605	+78%	0	0	—
case-17 I got a 'Not configured' error from Observal. I need to set up a user-scoped fallback agent named 'tutor' for OpenCode. What file path should be created? Format as a one-sentence intent, fenced code block, result, and next step.	fail→pass	5,995	5,016	-16%	1	1	0%	1,234	1,918	+55%	0	0	—
case-18 Observal returned 'Connection failed'. I want to set up a user-level fallback agent named 'copilot-assistant' for GitHub Copilot CLI. What path is expected? Structure your answer as a one-sentence intent statement, code block, result outcome, and next action.	fail→pass	5,239	4,407	-16%	1	1	0%	1,022	1,702	+67%	0	0	—
case-19 Observal CLI produced 'Not configured'. I need to create a project-scoped fallback agent named 'coder' for Codex CLI. Where should this file be written? Format your response with a single-sentence intent summary, code fence, result description, and next step.	fail→pass	4,955	4,062	-18%	1	1	0%	978	1,615	+65%	0	0	—
case-20 I am writing a Python script and want to initialize the Observal Python SDK tracing handler in my code to capture function spans automatically. How do I configure this trace wrapper in Python? Structure the output as a one-sentence intent statement, code block, result status, and next step or 'done'.	pass→pass	4,191	4,210	+0%	1	1	0%	895	1,622	+81%	0	0	—
case-21 I am administering the Observal backend server and need to run PostgreSQL database migrations for the telemetry storage schema. What command handles server schema migration? Format your output with a one-sentence goal statement, code block, expected result, and next step or 'done'.	fail→pass	4,437	7,516	+69%	1	1	0%	830	2,319	+179%	0	0	—
case-22 The Observal server is online and reachable, and I want to execute an online agent named 'deployer' with a prompt 'Deploy to staging'. What is the standard online CLI command to run this agent? Format your response as a one-sentence intent summary, code block, result status, and next step or 'done'.	pass→pass	3,082	3,048	-1%	1	1	0%	648	1,339	+107%	0	0	—

These cases were also screened by the skill’s intent / safety pass.See the Trust review ↗Reproduce this benchmark in Colab ↗

How this skill was measured

DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +86 percentage points is the difference between those two pass rates over the 22 comparable cases.

Model measured: gemini-3.6-flash
Graded by: gemini-3.6-flash
Grading method: judged
Verification: verified
Pass rate without the skill: 13.6%
Pass rate with the skill: 100%
Cases attempted: 22
Cases counted toward lift: 22
Cases that regressed: 0
Skill version measured: v1
Run completed: 8/8/2026, 16:19 UTC

Measured example — same prompt, with vs. without the skill · from the benchmark run · case case-01

Prompt

My shell crashed mid-process and I need to manually upload my local JSONL session logs to the server so trace context is complete. What is the execution step to reconcile my sessions? Structure the output by first stating your goal in one sentence, then giving the exact terminal command in a code fence, followed by the result status, and ending with the next action or 'done'.

Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.

Show both answers

SkillScore blends 4 signals1 of 4 complete

BenchmarkPass rate WITH skill on a verified run100

Live evalNeeds ≥10 evaluated traces in 30 days

AI ratingNeeds ≥10 rated traces

AdoptionNeeds ≥10 serves across ≥2 orgs in 30 days

Efficiency · separate axis (at equal correctness)+0% turns+77% tokens

Shown for transparency; not credited unless a verified run shows no correctness regression.

Backed by 1 of 4 signals — each added signal makes the score more robust.

Verified on gemini-3.6-flash · re-tested 50 days ago

Other measured skills in the registry, with their headline benchmark lift.